Contextual online advertising with user data from U.S. Census¶
Scenario¶
Consider an ad server that knows all of the information above about a user, except the education level. On the other hand, the ad network is managing ads that address a specific education level. For example, at any given time, the ad server has one ad that is targeting users with college education, one ad that is targeting users with elementary school education etc. If the target audience of the ad that is shown to a user matches the user's education level, there is a high probability of a click. If not, the probability of a click decreases gradually as the discrepancy between the target education level and the user's education level. In other words, the ad server is implicitly trying to predict users' education levels as close as possible.
In [ ]:
from collections import namedtuple
from numpy.random import uniform as U
import pandas as pd
import numpy as np
import io
import requests
from tensorflow import keras
from tensorflow.keras.layers import Dense, Dropout
Data¶
We will use a dataset that was modified from the 1994 U.S. Census and adapt it to an online advertising setting. The dataset is known as "Census Income Dataset" and available at https://archive.ics.uci.edu/ml/datasets/Census+Income.
In this dataset we use the following information on the individuals participated in the census: Age, work class, education, marital status, occupation, relationship, race, gender, work hours per week, native country, and income level.
With that, let's discuss how to turn this data into an online advertising scenario.
In [ ]:
url="https://archive.ics.uci.edu/ml/machine-learning-databases/adult/adult.data"
s=requests.get(url).content
names = ['age',
'workclass',
'fnlwgt',
'education',
'education_num',
'marital_status',
'occupation',
'relationship',
'race',
'gender',
'capital_gain',
'capital_loss',
'hours_per_week',
'native_country',
'income']
usecols = ['age',
'workclass',
'education',
'marital_status',
'occupation',
'relationship',
'race',
'gender',
'hours_per_week',
'native_country',
'income']
df_census = pd.read_csv(io.StringIO(s.decode('utf-8')),
sep=',',
skipinitialspace=True,
names=names,
header=None,
usecols=usecols)
Next, let's prepare the data set for our scenario.
In [ ]:
# Cleanup
df_census = df_census.replace('?', np.nan).dropna()
edu_map = {'Preschool': 'Elementary',
'1st-4th': 'Elementary',
'5th-6th': 'Elementary',
'7th-8th': 'Elementary',
'9th': 'Middle',
'10th': 'Middle',
'11th': 'Middle',
'12th': 'Middle',
'Some-college': 'Undergraduate',
'Bachelors': 'Undergraduate',
'Assoc-acdm': 'Undergraduate',
'Assoc-voc': 'Undergraduate',
'Prof-school': 'Graduate',
'Masters': 'Graduate',
'Doctorate': 'Graduate'}
for from_level, to_level in edu_map.items():
df_census.education.replace(from_level, to_level, inplace=True)
# Convert raw data to processed data
context_cols = [c for c in usecols if c != 'education']
df_data = pd.concat([pd.get_dummies(df_census[context_cols]),
df_census['education']], axis=1)
By doing this conversion at the beginning, we assume that we know all possible work class categories, native countries etc.
That's it! We have the data ready. Next, we implement a logic to simulate an ad click based on the actual education level of the user and the education level the ad displayed targets.
Simulating ad clicks¶
In this example, the availability of ads is also random in addition to the ad clicks being stochastic. We need to come up with some logic to simulate this behavior.
Let's start with determining the ad availability probabilities for each education category and implement the sampling of ads.
In [ ]:
def get_ad_inventory():
ad_inv_prob = {'Elementary': 0.9,
'Middle': 0.7,
'HS-grad': 0.7,
'Undergraduate': 0.9,
'Graduate': 0.8}
ad_inventory = []
for level, prob in ad_inv_prob.items():
if U() < prob:
ad_inventory.append(level)
# Make sure there are at least one ad
if not ad_inventory:
ad_inventory = get_ad_inventory()
return ad_inventory
As mentioned above, the ad server will have at most one ad for each target group. We also ensure that there is at least one ad in the inventory.
Then, we define a function to generate a click probabilistically, where the likelihood of a click increases to the degree that the user's education level and the ad's target match.
In [ ]:
def get_ad_click_probs():
base_prob = 0.8
delta = 0.3
ed_levels = {'Elementary': 1,
'Middle': 2,
'HS-grad': 3,
'Undergraduate': 4,
'Graduate': 5}
ad_click_probs = {l1: {l2: max(0, base_prob - delta * abs(ed_levels[l1]- ed_levels[l2])) for l2 in ed_levels}
for l1 in ed_levels}
return ad_click_probs
In [ ]:
def display_ad(ad_click_probs, user, ad):
prob = ad_click_probs[ad][user['education']]
click = 1 if U() < prob else 0
return click
So, when an ad is shown to a user, if the ad's target matches the user's education level, there will be chance of a click. This probability decreases by for each level of mismatch. For example, a person with a high school diploma has a chance to click on an ad that targets a user group of elementary school graduates (or college graduates). Note that this information is not known to the CB algorithm. It will be used only to simulate the clicks.
We have the problem set up. Next, we turn to implementing a CB model.
In [ ]:
def calc_regret(user, ad_inventory, ad_click_probs, ad_selected):
this_p = 0
max_p = 0
for ad in ad_inventory:
p = ad_click_probs[ad][user['education']]
if ad == ad_selected:
this_p = p
if p > max_p:
max_p = p
regret = max_p - this_p
return regret
In [ ]:
def get_model(n_input, dropout):
inputs = keras.Input(shape=(n_input,))
x = Dense(256, activation='relu')(inputs)
if dropout > 0:
x = Dropout(dropout)(x, training=True)
x = Dense(256, activation='relu')(x)
if dropout > 0:
x = Dropout(dropout)(x, training=True)
phat = Dense(1, activation='sigmoid')(x)
model = keras.Model(inputs, phat)
model.compile(loss=keras.losses.BinaryCrossentropy(),
optimizer=keras.optimizers.Adam(),
metrics=[keras.metrics.binary_accuracy])
return model
In [ ]:
def update_model(model, X, y):
X = np.array(X)
X = X.reshape((X.shape[0], X.shape[2]))
y = np.array(y).reshape(-1)
model.fit(X, y, epochs=10)
return model
In [ ]:
def ad_to_one_hot(ad):
ed_levels = ['Elementary',
'Middle',
'HS-grad',
'Undergraduate',
'Graduate']
ad_input = [0] * len(ed_levels)
if ad in ed_levels:
ad_input[ed_levels.index(ad)] = 1
return ad_input
In [ ]:
def select_ad(model, context, ad_inventory):
selected_ad = None
selected_x = None
max_action_val = 0
for ad in ad_inventory:
ad_x = ad_to_one_hot(ad)
x = np.array(context + ad_x).reshape((1, -1))
action_val_pred = model.predict(x)[0][0]
if action_val_pred >= max_action_val:
selected_ad = ad
selected_x = x
max_action_val = action_val_pred
return selected_ad, selected_x
In [ ]:
def generate_user(df_data):
user = df_data.sample(1)
context = user.iloc[:, :-1].values.tolist()[0]
return user.to_dict(orient='records')[0], context
In [ ]:
#collapse-hide
ad_click_probs = get_ad_click_probs()
df_cbandits = pd.DataFrame()
dropout_levels = [0, 0.01, 0.05, 0.1, 0.2, 0.4]
for d in dropout_levels:
print("Trying with dropout:", d)
np.random.seed(0)
context_n = df_data.shape[1] - 1
ad_input_n = df_data.education.nunique()
model = get_model(context_n + ad_input_n, 0.01)
X = []
y = []
regret_vec = []
total_regret = 0
for i in range(5000):
if i % 20 == 0:
print("# of impressions:", i)
user, context = generate_user(df_data)
ad_inventory = get_ad_inventory()
ad, x = select_ad(model, context, ad_inventory)
click = display_ad(ad_click_probs, user, ad)
regret = calc_regret(user, ad_inventory, ad_click_probs, ad)
total_regret += regret
regret_vec.append(total_regret)
X.append(x)
y.append(click)
if (i + 1) % 500 == 0:
print('Updating the model at', i+1)
model = update_model(model, X, y)
X = []
y = []
df_cbandits['dropout: '+str(d)] = regret_vec
Trying with dropout: 0 # of impressions: 0 # of impressions: 20 # of impressions: 40 # of impressions: 60 # of impressions: 80 # of impressions: 100 # of impressions: 120 # of impressions: 140 # of impressions: 160 # of impressions: 180 # of impressions: 200 # of impressions: 220 # of impressions: 240 # of impressions: 260 # of impressions: 280 # of impressions: 300 # of impressions: 320 # of impressions: 340 # of impressions: 360 # of impressions: 380 # of impressions: 400 # of impressions: 420 # of impressions: 440 # of impressions: 460 # of impressions: 480 Updating the model at 500 Epoch 1/10 16/16 [==============================] - 1s 3ms/step - loss: 0.8424 - binary_accuracy: 0.5600 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7275 - binary_accuracy: 0.5700 Epoch 3/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6772 - binary_accuracy: 0.6280 Epoch 4/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6462 - binary_accuracy: 0.6560 Epoch 5/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6463 - binary_accuracy: 0.6440 Epoch 6/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6442 - binary_accuracy: 0.6540 Epoch 7/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6532 - binary_accuracy: 0.6560 Epoch 8/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6247 - binary_accuracy: 0.6620 Epoch 9/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6254 - binary_accuracy: 0.6800 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6935 - binary_accuracy: 0.6100 # of impressions: 500 # of impressions: 520 # of impressions: 540 # of impressions: 560 # of impressions: 580 # of impressions: 600 # of impressions: 620 # of impressions: 640 # of impressions: 660 # of impressions: 680 # of impressions: 700 # of impressions: 720 # of impressions: 740 # of impressions: 760 # of impressions: 780 # of impressions: 800 # of impressions: 820 # of impressions: 840 # of impressions: 860 # of impressions: 880 # of impressions: 900 # of impressions: 920 # of impressions: 940 # of impressions: 960 # of impressions: 980 Updating the model at 1000 Epoch 1/10 16/16 [==============================] - 0s 2ms/step - loss: 0.7602 - binary_accuracy: 0.5100 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.8014 - binary_accuracy: 0.4860 Epoch 3/10 16/16 [==============================] - 0s 2ms/step - loss: 0.7363 - binary_accuracy: 0.5600 Epoch 4/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6868 - binary_accuracy: 0.5860 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6865 - binary_accuracy: 0.5420 Epoch 6/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6697 - binary_accuracy: 0.6020 Epoch 7/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6742 - binary_accuracy: 0.5760 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6855 - binary_accuracy: 0.5700 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6727 - binary_accuracy: 0.5840 Epoch 10/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6695 - binary_accuracy: 0.5940 # of impressions: 1000 # of impressions: 1020 # of impressions: 1040 # of impressions: 1060 # of impressions: 1080 # of impressions: 1100 # of impressions: 1120 # of impressions: 1140 # of impressions: 1160 # of impressions: 1180 # of impressions: 1200 # of impressions: 1220 # of impressions: 1240 # of impressions: 1260 # of impressions: 1280 # of impressions: 1300 # of impressions: 1320 # of impressions: 1340 # of impressions: 1360 # of impressions: 1380 # of impressions: 1400 # of impressions: 1420 # of impressions: 1440 # of impressions: 1460 # of impressions: 1480 Updating the model at 1500 Epoch 1/10 16/16 [==============================] - 0s 2ms/step - loss: 0.7112 - binary_accuracy: 0.4980 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7090 - binary_accuracy: 0.5460 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6822 - binary_accuracy: 0.5620 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6733 - binary_accuracy: 0.5980 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6872 - binary_accuracy: 0.5840 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6832 - binary_accuracy: 0.5740 Epoch 7/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6683 - binary_accuracy: 0.6000 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7067 - binary_accuracy: 0.5480 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6641 - binary_accuracy: 0.6240 Epoch 10/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6753 - binary_accuracy: 0.6000 # of impressions: 1500 # of impressions: 1520 # of impressions: 1540 # of impressions: 1560 # of impressions: 1580 # of impressions: 1600 # of impressions: 1620 # of impressions: 1640 # of impressions: 1660 # of impressions: 1680 # of impressions: 1700 # of impressions: 1720 # of impressions: 1740 # of impressions: 1760 # of impressions: 1780 # of impressions: 1800 # of impressions: 1820 # of impressions: 1840 # of impressions: 1860 # of impressions: 1880 # of impressions: 1900 # of impressions: 1920 # of impressions: 1940 # of impressions: 1960 # of impressions: 1980 Updating the model at 2000 Epoch 1/10 16/16 [==============================] - 0s 2ms/step - loss: 0.7047 - binary_accuracy: 0.5480 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6834 - binary_accuracy: 0.5920 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6927 - binary_accuracy: 0.5640 Epoch 4/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6886 - binary_accuracy: 0.5700 Epoch 5/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6585 - binary_accuracy: 0.6100 Epoch 6/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6880 - binary_accuracy: 0.5840 Epoch 7/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6924 - binary_accuracy: 0.5900 Epoch 8/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6759 - binary_accuracy: 0.5760 Epoch 9/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6593 - binary_accuracy: 0.6340 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6594 - binary_accuracy: 0.5740 # of impressions: 2000 # of impressions: 2020 # of impressions: 2040 # of impressions: 2060 # of impressions: 2080 # of impressions: 2100 # of impressions: 2120 # of impressions: 2140 # of impressions: 2160 # of impressions: 2180 # of impressions: 2200 # of impressions: 2220 # of impressions: 2240 # of impressions: 2260 # of impressions: 2280 # of impressions: 2300 # of impressions: 2320 # of impressions: 2340 # of impressions: 2360 # of impressions: 2380 # of impressions: 2400 # of impressions: 2420 # of impressions: 2440 # of impressions: 2460 # of impressions: 2480 Updating the model at 2500 Epoch 1/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6748 - binary_accuracy: 0.6140 Epoch 2/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6671 - binary_accuracy: 0.6380 Epoch 3/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6687 - binary_accuracy: 0.6200 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6667 - binary_accuracy: 0.6220 Epoch 5/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6570 - binary_accuracy: 0.6400 Epoch 6/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6470 - binary_accuracy: 0.6320 Epoch 7/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6573 - binary_accuracy: 0.6240 Epoch 8/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6706 - binary_accuracy: 0.6380 Epoch 9/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6372 - binary_accuracy: 0.6600 Epoch 10/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6384 - binary_accuracy: 0.6500 # of impressions: 2500 # of impressions: 2520 # of impressions: 2540 # of impressions: 2560 # of impressions: 2580 # of impressions: 2600 # of impressions: 2620 # of impressions: 2640 # of impressions: 2660 # of impressions: 2680 # of impressions: 2700 # of impressions: 2720 # of impressions: 2740 # of impressions: 2760 # of impressions: 2780 # of impressions: 2800 # of impressions: 2820 # of impressions: 2840 # of impressions: 2860 # of impressions: 2880 # of impressions: 2900 # of impressions: 2920 # of impressions: 2940 # of impressions: 2960 # of impressions: 2980 Updating the model at 3000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7047 - binary_accuracy: 0.5280 Epoch 2/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6950 - binary_accuracy: 0.6000 Epoch 3/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6773 - binary_accuracy: 0.5880 Epoch 4/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6734 - binary_accuracy: 0.5920 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6644 - binary_accuracy: 0.5960 Epoch 6/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6815 - binary_accuracy: 0.5880 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6714 - binary_accuracy: 0.5920 Epoch 8/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6674 - binary_accuracy: 0.6000 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6491 - binary_accuracy: 0.6460 Epoch 10/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6584 - binary_accuracy: 0.6120 # of impressions: 3000 # of impressions: 3020 # of impressions: 3040 # of impressions: 3060 # of impressions: 3080 # of impressions: 3100 # of impressions: 3120 # of impressions: 3140 # of impressions: 3160 # of impressions: 3180 # of impressions: 3200 # of impressions: 3220 # of impressions: 3240 # of impressions: 3260 # of impressions: 3280 # of impressions: 3300 # of impressions: 3320 # of impressions: 3340 # of impressions: 3360 # of impressions: 3380 # of impressions: 3400 # of impressions: 3420 # of impressions: 3440 # of impressions: 3460 # of impressions: 3480 Updating the model at 3500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7192 - binary_accuracy: 0.5240 Epoch 2/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6999 - binary_accuracy: 0.5240 Epoch 3/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6960 - binary_accuracy: 0.5520 Epoch 4/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6843 - binary_accuracy: 0.5440 Epoch 5/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6735 - binary_accuracy: 0.5940 Epoch 6/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6647 - binary_accuracy: 0.5920 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6557 - binary_accuracy: 0.5900 Epoch 8/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6777 - binary_accuracy: 0.5760 Epoch 9/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6630 - binary_accuracy: 0.5920 Epoch 10/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6650 - binary_accuracy: 0.5980 # of impressions: 3500 # of impressions: 3520 # of impressions: 3540 # of impressions: 3560 # of impressions: 3580 # of impressions: 3600 # of impressions: 3620 # of impressions: 3640 # of impressions: 3660 # of impressions: 3680 # of impressions: 3700 # of impressions: 3720 # of impressions: 3740 # of impressions: 3760 # of impressions: 3780 # of impressions: 3800 # of impressions: 3820 # of impressions: 3840 # of impressions: 3860 # of impressions: 3880 # of impressions: 3900 # of impressions: 3920 # of impressions: 3940 # of impressions: 3960 # of impressions: 3980 Updating the model at 4000 Epoch 1/10 16/16 [==============================] - 0s 2ms/step - loss: 0.7114 - binary_accuracy: 0.5680 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7233 - binary_accuracy: 0.5420 Epoch 3/10 16/16 [==============================] - 0s 2ms/step - loss: 0.7212 - binary_accuracy: 0.4840 Epoch 4/10 16/16 [==============================] - 0s 2ms/step - loss: 0.7029 - binary_accuracy: 0.5300 Epoch 5/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6793 - binary_accuracy: 0.5960 Epoch 6/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6729 - binary_accuracy: 0.6140 Epoch 7/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6804 - binary_accuracy: 0.5740 Epoch 8/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6612 - binary_accuracy: 0.5900 Epoch 9/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6541 - binary_accuracy: 0.6220 Epoch 10/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6443 - binary_accuracy: 0.6120 # of impressions: 4000 # of impressions: 4020 # of impressions: 4040 # of impressions: 4060 # of impressions: 4080 # of impressions: 4100 # of impressions: 4120 # of impressions: 4140 # of impressions: 4160 # of impressions: 4180 # of impressions: 4200 # of impressions: 4220 # of impressions: 4240 # of impressions: 4260 # of impressions: 4280 # of impressions: 4300 # of impressions: 4320 # of impressions: 4340 # of impressions: 4360 # of impressions: 4380 # of impressions: 4400 # of impressions: 4420 # of impressions: 4440 # of impressions: 4460 # of impressions: 4480 Updating the model at 4500 Epoch 1/10 16/16 [==============================] - 0s 2ms/step - loss: 0.7017 - binary_accuracy: 0.5300 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6915 - binary_accuracy: 0.5300 Epoch 3/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6861 - binary_accuracy: 0.5660 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6814 - binary_accuracy: 0.5720 Epoch 5/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6816 - binary_accuracy: 0.5900 Epoch 6/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6762 - binary_accuracy: 0.5940 Epoch 7/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6685 - binary_accuracy: 0.5960 Epoch 8/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6679 - binary_accuracy: 0.6160 Epoch 9/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6638 - binary_accuracy: 0.6180 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6589 - binary_accuracy: 0.6040 # of impressions: 4500 # of impressions: 4520 # of impressions: 4540 # of impressions: 4560 # of impressions: 4580 # of impressions: 4600 # of impressions: 4620 # of impressions: 4640 # of impressions: 4660 # of impressions: 4680 # of impressions: 4700 # of impressions: 4720 # of impressions: 4740 # of impressions: 4760 # of impressions: 4780 # of impressions: 4800 # of impressions: 4820 # of impressions: 4840 # of impressions: 4860 # of impressions: 4880 # of impressions: 4900 # of impressions: 4920 # of impressions: 4940 # of impressions: 4960 # of impressions: 4980 Updating the model at 5000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6787 - binary_accuracy: 0.6060 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6713 - binary_accuracy: 0.6220 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6632 - binary_accuracy: 0.6220 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6589 - binary_accuracy: 0.6240 Epoch 5/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6574 - binary_accuracy: 0.6320 Epoch 6/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6572 - binary_accuracy: 0.6160 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6503 - binary_accuracy: 0.6300 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6458 - binary_accuracy: 0.6380 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6434 - binary_accuracy: 0.6400 Epoch 10/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6428 - binary_accuracy: 0.6400 Trying with dropout: 0.01 # of impressions: 0 # of impressions: 20 # of impressions: 40 # of impressions: 60 # of impressions: 80 # of impressions: 100 # of impressions: 120 # of impressions: 140 # of impressions: 160 # of impressions: 180 # of impressions: 200 # of impressions: 220 # of impressions: 240 # of impressions: 260 # of impressions: 280 # of impressions: 300 # of impressions: 320 # of impressions: 340 # of impressions: 360 # of impressions: 380 # of impressions: 400 # of impressions: 420 # of impressions: 440 # of impressions: 460 # of impressions: 480 Updating the model at 500 Epoch 1/10 16/16 [==============================] - 1s 3ms/step - loss: 1.5034 - binary_accuracy: 0.5400 Epoch 2/10 16/16 [==============================] - 0s 2ms/step - loss: 0.7648 - binary_accuracy: 0.5500 Epoch 3/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6662 - binary_accuracy: 0.5960 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7367 - binary_accuracy: 0.5400 Epoch 5/10 16/16 [==============================] - 0s 2ms/step - loss: 0.7202 - binary_accuracy: 0.5600 Epoch 6/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6967 - binary_accuracy: 0.5980 Epoch 7/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6763 - binary_accuracy: 0.5960 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7087 - binary_accuracy: 0.5440 Epoch 9/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6692 - binary_accuracy: 0.6100 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7623 - binary_accuracy: 0.5580 # of impressions: 500 # of impressions: 520 # of impressions: 540 # of impressions: 560 # of impressions: 580 # of impressions: 600 # of impressions: 620 # of impressions: 640 # of impressions: 660 # of impressions: 680 # of impressions: 700 # of impressions: 720 # of impressions: 740 # of impressions: 760 # of impressions: 780 # of impressions: 800 # of impressions: 820 # of impressions: 840 # of impressions: 860 # of impressions: 880 # of impressions: 900 # of impressions: 920 # of impressions: 940 # of impressions: 960 # of impressions: 980 Updating the model at 1000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7131 - binary_accuracy: 0.5240 Epoch 2/10 16/16 [==============================] - 0s 2ms/step - loss: 0.7190 - binary_accuracy: 0.5260 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7115 - binary_accuracy: 0.5520 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6994 - binary_accuracy: 0.5300 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7226 - binary_accuracy: 0.5380 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7089 - binary_accuracy: 0.5420 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6813 - binary_accuracy: 0.5640 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6568 - binary_accuracy: 0.6100 Epoch 9/10 16/16 [==============================] - 0s 2ms/step - loss: 0.7108 - binary_accuracy: 0.5700 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7122 - binary_accuracy: 0.5660 # of impressions: 1000 # of impressions: 1020 # of impressions: 1040 # of impressions: 1060 # of impressions: 1080 # of impressions: 1100 # of impressions: 1120 # of impressions: 1140 # of impressions: 1160 # of impressions: 1180 # of impressions: 1200 # of impressions: 1220 # of impressions: 1240 # of impressions: 1260 # of impressions: 1280 # of impressions: 1300 # of impressions: 1320 # of impressions: 1340 # of impressions: 1360 # of impressions: 1380 # of impressions: 1400 # of impressions: 1420 # of impressions: 1440 # of impressions: 1460 # of impressions: 1480 Updating the model at 1500 Epoch 1/10 16/16 [==============================] - 0s 2ms/step - loss: 0.7689 - binary_accuracy: 0.5080 Epoch 2/10 16/16 [==============================] - 0s 2ms/step - loss: 0.7372 - binary_accuracy: 0.5600 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6954 - binary_accuracy: 0.5660 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6958 - binary_accuracy: 0.5720 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6695 - binary_accuracy: 0.5880 Epoch 6/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6546 - binary_accuracy: 0.6360 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6767 - binary_accuracy: 0.5960 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6611 - binary_accuracy: 0.6040 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6645 - binary_accuracy: 0.6140 Epoch 10/10 16/16 [==============================] - 0s 2ms/step - loss: 0.7341 - binary_accuracy: 0.5540 # of impressions: 1500 # of impressions: 1520 # of impressions: 1540 # of impressions: 1560 # of impressions: 1580 # of impressions: 1600 # of impressions: 1620 # of impressions: 1640 # of impressions: 1660 # of impressions: 1680 # of impressions: 1700 # of impressions: 1720 # of impressions: 1740 # of impressions: 1760 # of impressions: 1780 # of impressions: 1800 # of impressions: 1820 # of impressions: 1840 # of impressions: 1860 # of impressions: 1880 # of impressions: 1900 # of impressions: 1920 # of impressions: 1940 # of impressions: 1960 # of impressions: 1980 Updating the model at 2000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7322 - binary_accuracy: 0.5600 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7008 - binary_accuracy: 0.5800 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6948 - binary_accuracy: 0.5680 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6803 - binary_accuracy: 0.5780 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6858 - binary_accuracy: 0.5920 Epoch 6/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6883 - binary_accuracy: 0.5560 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6837 - binary_accuracy: 0.5940 Epoch 8/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6399 - binary_accuracy: 0.6280 Epoch 9/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6560 - binary_accuracy: 0.6340 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6643 - binary_accuracy: 0.5980 # of impressions: 2000 # of impressions: 2020 # of impressions: 2040 # of impressions: 2060 # of impressions: 2080 # of impressions: 2100 # of impressions: 2120 # of impressions: 2140 # of impressions: 2160 # of impressions: 2180 # of impressions: 2200 # of impressions: 2220 # of impressions: 2240 # of impressions: 2260 # of impressions: 2280 # of impressions: 2300 # of impressions: 2320 # of impressions: 2340 # of impressions: 2360 # of impressions: 2380 # of impressions: 2400 # of impressions: 2420 # of impressions: 2440 # of impressions: 2460 # of impressions: 2480 Updating the model at 2500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6871 - binary_accuracy: 0.5940 Epoch 2/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6683 - binary_accuracy: 0.6340 Epoch 3/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6588 - binary_accuracy: 0.6400 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6966 - binary_accuracy: 0.5480 Epoch 5/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6862 - binary_accuracy: 0.6120 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6529 - binary_accuracy: 0.6420 Epoch 7/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6500 - binary_accuracy: 0.6520 Epoch 8/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6326 - binary_accuracy: 0.6520 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6464 - binary_accuracy: 0.6320 Epoch 10/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6301 - binary_accuracy: 0.6540 # of impressions: 2500 # of impressions: 2520 # of impressions: 2540 # of impressions: 2560 # of impressions: 2580 # of impressions: 2600 # of impressions: 2620 # of impressions: 2640 # of impressions: 2660 # of impressions: 2680 # of impressions: 2700 # of impressions: 2720 # of impressions: 2740 # of impressions: 2760 # of impressions: 2780 # of impressions: 2800 # of impressions: 2820 # of impressions: 2840 # of impressions: 2860 # of impressions: 2880 # of impressions: 2900 # of impressions: 2920 # of impressions: 2940 # of impressions: 2960 # of impressions: 2980 Updating the model at 3000 Epoch 1/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6894 - binary_accuracy: 0.5820 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6624 - binary_accuracy: 0.6040 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6543 - binary_accuracy: 0.6140 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6708 - binary_accuracy: 0.5960 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6855 - binary_accuracy: 0.5940 Epoch 6/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6892 - binary_accuracy: 0.6000 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6642 - binary_accuracy: 0.6040 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6752 - binary_accuracy: 0.5820 Epoch 9/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6389 - binary_accuracy: 0.6220 Epoch 10/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6285 - binary_accuracy: 0.6640 # of impressions: 3000 # of impressions: 3020 # of impressions: 3040 # of impressions: 3060 # of impressions: 3080 # of impressions: 3100 # of impressions: 3120 # of impressions: 3140 # of impressions: 3160 # of impressions: 3180 # of impressions: 3200 # of impressions: 3220 # of impressions: 3240 # of impressions: 3260 # of impressions: 3280 # of impressions: 3300 # of impressions: 3320 # of impressions: 3340 # of impressions: 3360 # of impressions: 3380 # of impressions: 3400 # of impressions: 3420 # of impressions: 3440 # of impressions: 3460 # of impressions: 3480 Updating the model at 3500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6938 - binary_accuracy: 0.6000 Epoch 2/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6976 - binary_accuracy: 0.5820 Epoch 3/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6909 - binary_accuracy: 0.5880 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6727 - binary_accuracy: 0.5840 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6619 - binary_accuracy: 0.6220 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6533 - binary_accuracy: 0.6040 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6569 - binary_accuracy: 0.6200 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6566 - binary_accuracy: 0.6340 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6500 - binary_accuracy: 0.6320 Epoch 10/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6476 - binary_accuracy: 0.6240 # of impressions: 3500 # of impressions: 3520 # of impressions: 3540 # of impressions: 3560 # of impressions: 3580 # of impressions: 3600 # of impressions: 3620 # of impressions: 3640 # of impressions: 3660 # of impressions: 3680 # of impressions: 3700 # of impressions: 3720 # of impressions: 3740 # of impressions: 3760 # of impressions: 3780 # of impressions: 3800 # of impressions: 3820 # of impressions: 3840 # of impressions: 3860 # of impressions: 3880 # of impressions: 3900 # of impressions: 3920 # of impressions: 3940 # of impressions: 3960 # of impressions: 3980 Updating the model at 4000 Epoch 1/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6862 - binary_accuracy: 0.5920 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6774 - binary_accuracy: 0.6360 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6600 - binary_accuracy: 0.6060 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6576 - binary_accuracy: 0.6200 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6465 - binary_accuracy: 0.6420 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6671 - binary_accuracy: 0.5920 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6541 - binary_accuracy: 0.6160 Epoch 8/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6515 - binary_accuracy: 0.6220 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6366 - binary_accuracy: 0.6400 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6327 - binary_accuracy: 0.6480 # of impressions: 4000 # of impressions: 4020 # of impressions: 4040 # of impressions: 4060 # of impressions: 4080 # of impressions: 4100 # of impressions: 4120 # of impressions: 4140 # of impressions: 4160 # of impressions: 4180 # of impressions: 4200 # of impressions: 4220 # of impressions: 4240 # of impressions: 4260 # of impressions: 4280 # of impressions: 4300 # of impressions: 4320 # of impressions: 4340 # of impressions: 4360 # of impressions: 4380 # of impressions: 4400 # of impressions: 4420 # of impressions: 4440 # of impressions: 4460 # of impressions: 4480 Updating the model at 4500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6963 - binary_accuracy: 0.5540 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6808 - binary_accuracy: 0.5780 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6785 - binary_accuracy: 0.5780 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6676 - binary_accuracy: 0.6020 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6620 - binary_accuracy: 0.6040 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6663 - binary_accuracy: 0.5780 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6683 - binary_accuracy: 0.5800 Epoch 8/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6672 - binary_accuracy: 0.5840 Epoch 9/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6665 - binary_accuracy: 0.5800 Epoch 10/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6581 - binary_accuracy: 0.6000 # of impressions: 4500 # of impressions: 4520 # of impressions: 4540 # of impressions: 4560 # of impressions: 4580 # of impressions: 4600 # of impressions: 4620 # of impressions: 4640 # of impressions: 4660 # of impressions: 4680 # of impressions: 4700 # of impressions: 4720 # of impressions: 4740 # of impressions: 4760 # of impressions: 4780 # of impressions: 4800 # of impressions: 4820 # of impressions: 4840 # of impressions: 4860 # of impressions: 4880 # of impressions: 4900 # of impressions: 4920 # of impressions: 4940 # of impressions: 4960 # of impressions: 4980 Updating the model at 5000 Epoch 1/10 16/16 [==============================] - 0s 2ms/step - loss: 0.7103 - binary_accuracy: 0.5280 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6982 - binary_accuracy: 0.5580 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6724 - binary_accuracy: 0.5960 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6622 - binary_accuracy: 0.6160 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6637 - binary_accuracy: 0.6060 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6502 - binary_accuracy: 0.6440 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6495 - binary_accuracy: 0.6200 Epoch 8/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6643 - binary_accuracy: 0.5860 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6513 - binary_accuracy: 0.6220 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6470 - binary_accuracy: 0.6320 Trying with dropout: 0.05 # of impressions: 0 # of impressions: 20 # of impressions: 40 # of impressions: 60 # of impressions: 80 # of impressions: 100 # of impressions: 120 # of impressions: 140 # of impressions: 160 # of impressions: 180 # of impressions: 200 # of impressions: 220 # of impressions: 240 # of impressions: 260 # of impressions: 280 # of impressions: 300 # of impressions: 320 # of impressions: 340 # of impressions: 360 # of impressions: 380 # of impressions: 400 # of impressions: 420 # of impressions: 440 # of impressions: 460 # of impressions: 480 Updating the model at 500 Epoch 1/10 16/16 [==============================] - 1s 3ms/step - loss: 0.8016 - binary_accuracy: 0.5700 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.8029 - binary_accuracy: 0.5700 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7046 - binary_accuracy: 0.6120 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6696 - binary_accuracy: 0.6200 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6876 - binary_accuracy: 0.5980 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6445 - binary_accuracy: 0.6280 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6531 - binary_accuracy: 0.6180 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6353 - binary_accuracy: 0.6460 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6844 - binary_accuracy: 0.6080 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6466 - binary_accuracy: 0.6560 # of impressions: 500 # of impressions: 520 # of impressions: 540 # of impressions: 560 # of impressions: 580 # of impressions: 600 # of impressions: 620 # of impressions: 640 # of impressions: 660 # of impressions: 680 # of impressions: 700 # of impressions: 720 # of impressions: 740 # of impressions: 760 # of impressions: 780 # of impressions: 800 # of impressions: 820 # of impressions: 840 # of impressions: 860 # of impressions: 880 # of impressions: 900 # of impressions: 920 # of impressions: 940 # of impressions: 960 # of impressions: 980 Updating the model at 1000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7171 - binary_accuracy: 0.5400 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7333 - binary_accuracy: 0.5060 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6868 - binary_accuracy: 0.5580 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6944 - binary_accuracy: 0.5540 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6704 - binary_accuracy: 0.5760 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6894 - binary_accuracy: 0.5620 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6757 - binary_accuracy: 0.5660 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6497 - binary_accuracy: 0.6080 Epoch 9/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6460 - binary_accuracy: 0.6240 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6662 - binary_accuracy: 0.5700 # of impressions: 1000 # of impressions: 1020 # of impressions: 1040 # of impressions: 1060 # of impressions: 1080 # of impressions: 1100 # of impressions: 1120 # of impressions: 1140 # of impressions: 1160 # of impressions: 1180 # of impressions: 1200 # of impressions: 1220 # of impressions: 1240 # of impressions: 1260 # of impressions: 1280 # of impressions: 1300 # of impressions: 1320 # of impressions: 1340 # of impressions: 1360 # of impressions: 1380 # of impressions: 1400 # of impressions: 1420 # of impressions: 1440 # of impressions: 1460 # of impressions: 1480 Updating the model at 1500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7188 - binary_accuracy: 0.5440 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6956 - binary_accuracy: 0.5680 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7030 - binary_accuracy: 0.5560 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6991 - binary_accuracy: 0.5640 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6999 - binary_accuracy: 0.5540 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6921 - binary_accuracy: 0.5700 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6708 - binary_accuracy: 0.5980 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6667 - binary_accuracy: 0.6000 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6702 - binary_accuracy: 0.6000 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6865 - binary_accuracy: 0.5700 # of impressions: 1500 # of impressions: 1520 # of impressions: 1540 # of impressions: 1560 # of impressions: 1580 # of impressions: 1600 # of impressions: 1620 # of impressions: 1640 # of impressions: 1660 # of impressions: 1680 # of impressions: 1700 # of impressions: 1720 # of impressions: 1740 # of impressions: 1760 # of impressions: 1780 # of impressions: 1800 # of impressions: 1820 # of impressions: 1840 # of impressions: 1860 # of impressions: 1880 # of impressions: 1900 # of impressions: 1920 # of impressions: 1940 # of impressions: 1960 # of impressions: 1980 Updating the model at 2000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7352 - binary_accuracy: 0.5580 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6909 - binary_accuracy: 0.5500 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7130 - binary_accuracy: 0.5740 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6912 - binary_accuracy: 0.5460 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6799 - binary_accuracy: 0.5900 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6799 - binary_accuracy: 0.5980 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6567 - binary_accuracy: 0.6140 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6618 - binary_accuracy: 0.6120 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6665 - binary_accuracy: 0.6060 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7091 - binary_accuracy: 0.5680 # of impressions: 2000 # of impressions: 2020 # of impressions: 2040 # of impressions: 2060 # of impressions: 2080 # of impressions: 2100 # of impressions: 2120 # of impressions: 2140 # of impressions: 2160 # of impressions: 2180 # of impressions: 2200 # of impressions: 2220 # of impressions: 2240 # of impressions: 2260 # of impressions: 2280 # of impressions: 2300 # of impressions: 2320 # of impressions: 2340 # of impressions: 2360 # of impressions: 2380 # of impressions: 2400 # of impressions: 2420 # of impressions: 2440 # of impressions: 2460 # of impressions: 2480 Updating the model at 2500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7086 - binary_accuracy: 0.5400 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6824 - binary_accuracy: 0.5800 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6741 - binary_accuracy: 0.6040 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6949 - binary_accuracy: 0.5620 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6811 - binary_accuracy: 0.5780 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6588 - binary_accuracy: 0.5980 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7057 - binary_accuracy: 0.5560 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7061 - binary_accuracy: 0.5580 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6653 - binary_accuracy: 0.6200 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6638 - binary_accuracy: 0.6020 # of impressions: 2500 # of impressions: 2520 # of impressions: 2540 # of impressions: 2560 # of impressions: 2580 # of impressions: 2600 # of impressions: 2620 # of impressions: 2640 # of impressions: 2660 # of impressions: 2680 # of impressions: 2700 # of impressions: 2720 # of impressions: 2740 # of impressions: 2760 # of impressions: 2780 # of impressions: 2800 # of impressions: 2820 # of impressions: 2840 # of impressions: 2860 # of impressions: 2880 # of impressions: 2900 # of impressions: 2920 # of impressions: 2940 # of impressions: 2960 # of impressions: 2980 Updating the model at 3000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6878 - binary_accuracy: 0.5820 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6770 - binary_accuracy: 0.5720 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6869 - binary_accuracy: 0.5700 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6680 - binary_accuracy: 0.5860 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6654 - binary_accuracy: 0.5980 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6540 - binary_accuracy: 0.6400 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6570 - binary_accuracy: 0.6220 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6701 - binary_accuracy: 0.6120 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6388 - binary_accuracy: 0.6360 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6397 - binary_accuracy: 0.6520 # of impressions: 3000 # of impressions: 3020 # of impressions: 3040 # of impressions: 3060 # of impressions: 3080 # of impressions: 3100 # of impressions: 3120 # of impressions: 3140 # of impressions: 3160 # of impressions: 3180 # of impressions: 3200 # of impressions: 3220 # of impressions: 3240 # of impressions: 3260 # of impressions: 3280 # of impressions: 3300 # of impressions: 3320 # of impressions: 3340 # of impressions: 3360 # of impressions: 3380 # of impressions: 3400 # of impressions: 3420 # of impressions: 3440 # of impressions: 3460 # of impressions: 3480 Updating the model at 3500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7046 - binary_accuracy: 0.5380 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6993 - binary_accuracy: 0.5480 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7059 - binary_accuracy: 0.5420 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6962 - binary_accuracy: 0.5480 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6715 - binary_accuracy: 0.5800 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6710 - binary_accuracy: 0.5800 Epoch 7/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6701 - binary_accuracy: 0.5800 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6563 - binary_accuracy: 0.6020 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6519 - binary_accuracy: 0.6060 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6578 - binary_accuracy: 0.6160 # of impressions: 3500 # of impressions: 3520 # of impressions: 3540 # of impressions: 3560 # of impressions: 3580 # of impressions: 3600 # of impressions: 3620 # of impressions: 3640 # of impressions: 3660 # of impressions: 3680 # of impressions: 3700 # of impressions: 3720 # of impressions: 3740 # of impressions: 3760 # of impressions: 3780 # of impressions: 3800 # of impressions: 3820 # of impressions: 3840 # of impressions: 3860 # of impressions: 3880 # of impressions: 3900 # of impressions: 3920 # of impressions: 3940 # of impressions: 3960 # of impressions: 3980 Updating the model at 4000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6873 - binary_accuracy: 0.5660 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6753 - binary_accuracy: 0.5680 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6653 - binary_accuracy: 0.5880 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6629 - binary_accuracy: 0.5880 Epoch 5/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6724 - binary_accuracy: 0.5780 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6657 - binary_accuracy: 0.6000 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6574 - binary_accuracy: 0.5760 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6437 - binary_accuracy: 0.6240 Epoch 9/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6496 - binary_accuracy: 0.6100 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6472 - binary_accuracy: 0.6380 # of impressions: 4000 # of impressions: 4020 # of impressions: 4040 # of impressions: 4060 # of impressions: 4080 # of impressions: 4100 # of impressions: 4120 # of impressions: 4140 # of impressions: 4160 # of impressions: 4180 # of impressions: 4200 # of impressions: 4220 # of impressions: 4240 # of impressions: 4260 # of impressions: 4280 # of impressions: 4300 # of impressions: 4320 # of impressions: 4340 # of impressions: 4360 # of impressions: 4380 # of impressions: 4400 # of impressions: 4420 # of impressions: 4440 # of impressions: 4460 # of impressions: 4480 Updating the model at 4500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7081 - binary_accuracy: 0.5620 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6928 - binary_accuracy: 0.5540 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6886 - binary_accuracy: 0.5860 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6752 - binary_accuracy: 0.5760 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6925 - binary_accuracy: 0.5640 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6689 - binary_accuracy: 0.5700 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6720 - binary_accuracy: 0.5740 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6671 - binary_accuracy: 0.5940 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6657 - binary_accuracy: 0.5920 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6640 - binary_accuracy: 0.6080 # of impressions: 4500 # of impressions: 4520 # of impressions: 4540 # of impressions: 4560 # of impressions: 4580 # of impressions: 4600 # of impressions: 4620 # of impressions: 4640 # of impressions: 4660 # of impressions: 4680 # of impressions: 4700 # of impressions: 4720 # of impressions: 4740 # of impressions: 4760 # of impressions: 4780 # of impressions: 4800 # of impressions: 4820 # of impressions: 4840 # of impressions: 4860 # of impressions: 4880 # of impressions: 4900 # of impressions: 4920 # of impressions: 4940 # of impressions: 4960 # of impressions: 4980 Updating the model at 5000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6828 - binary_accuracy: 0.5960 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6700 - binary_accuracy: 0.5820 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6599 - binary_accuracy: 0.6040 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6520 - binary_accuracy: 0.6240 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6579 - binary_accuracy: 0.6220 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6452 - binary_accuracy: 0.6380 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6448 - binary_accuracy: 0.6220 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6395 - binary_accuracy: 0.6420 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6481 - binary_accuracy: 0.6020 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6379 - binary_accuracy: 0.6520 Trying with dropout: 0.1 # of impressions: 0 # of impressions: 20 # of impressions: 40 # of impressions: 60 # of impressions: 80 # of impressions: 100 # of impressions: 120 # of impressions: 140 # of impressions: 160 # of impressions: 180 # of impressions: 200 # of impressions: 220 # of impressions: 240 # of impressions: 260 # of impressions: 280 # of impressions: 300 # of impressions: 320 # of impressions: 340 # of impressions: 360 # of impressions: 380 # of impressions: 400 # of impressions: 420 # of impressions: 440 # of impressions: 460 # of impressions: 480 Updating the model at 500 Epoch 1/10 16/16 [==============================] - 1s 3ms/step - loss: 0.9884 - binary_accuracy: 0.4940 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7877 - binary_accuracy: 0.5840 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6771 - binary_accuracy: 0.5860 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6991 - binary_accuracy: 0.5820 Epoch 5/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6906 - binary_accuracy: 0.5980 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7296 - binary_accuracy: 0.5380 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6380 - binary_accuracy: 0.6400 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6671 - binary_accuracy: 0.6140 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6816 - binary_accuracy: 0.6280 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6397 - binary_accuracy: 0.6280 # of impressions: 500 # of impressions: 520 # of impressions: 540 # of impressions: 560 # of impressions: 580 # of impressions: 600 # of impressions: 620 # of impressions: 640 # of impressions: 660 # of impressions: 680 # of impressions: 700 # of impressions: 720 # of impressions: 740 # of impressions: 760 # of impressions: 780 # of impressions: 800 # of impressions: 820 # of impressions: 840 # of impressions: 860 # of impressions: 880 # of impressions: 900 # of impressions: 920 # of impressions: 940 # of impressions: 960 # of impressions: 980 Updating the model at 1000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7042 - binary_accuracy: 0.5280 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7512 - binary_accuracy: 0.5220 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6795 - binary_accuracy: 0.5680 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6938 - binary_accuracy: 0.5640 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7466 - binary_accuracy: 0.5640 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6878 - binary_accuracy: 0.5840 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6762 - binary_accuracy: 0.5700 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6555 - binary_accuracy: 0.6160 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6798 - binary_accuracy: 0.5720 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6696 - binary_accuracy: 0.5760 # of impressions: 1000 # of impressions: 1020 # of impressions: 1040 # of impressions: 1060 # of impressions: 1080 # of impressions: 1100 # of impressions: 1120 # of impressions: 1140 # of impressions: 1160 # of impressions: 1180 # of impressions: 1200 # of impressions: 1220 # of impressions: 1240 # of impressions: 1260 # of impressions: 1280 # of impressions: 1300 # of impressions: 1320 # of impressions: 1340 # of impressions: 1360 # of impressions: 1380 # of impressions: 1400 # of impressions: 1420 # of impressions: 1440 # of impressions: 1460 # of impressions: 1480 Updating the model at 1500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7740 - binary_accuracy: 0.5060 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.8210 - binary_accuracy: 0.5000 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6919 - binary_accuracy: 0.5800 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7914 - binary_accuracy: 0.5460 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6909 - binary_accuracy: 0.5520 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6774 - binary_accuracy: 0.5820 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6563 - binary_accuracy: 0.6180 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6597 - binary_accuracy: 0.6000 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6667 - binary_accuracy: 0.5960 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6701 - binary_accuracy: 0.5840 # of impressions: 1500 # of impressions: 1520 # of impressions: 1540 # of impressions: 1560 # of impressions: 1580 # of impressions: 1600 # of impressions: 1620 # of impressions: 1640 # of impressions: 1660 # of impressions: 1680 # of impressions: 1700 # of impressions: 1720 # of impressions: 1740 # of impressions: 1760 # of impressions: 1780 # of impressions: 1800 # of impressions: 1820 # of impressions: 1840 # of impressions: 1860 # of impressions: 1880 # of impressions: 1900 # of impressions: 1920 # of impressions: 1940 # of impressions: 1960 # of impressions: 1980 Updating the model at 2000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7469 - binary_accuracy: 0.5380 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6965 - binary_accuracy: 0.5600 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6811 - binary_accuracy: 0.5820 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6811 - binary_accuracy: 0.6100 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6682 - binary_accuracy: 0.5940 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6606 - binary_accuracy: 0.6100 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6767 - binary_accuracy: 0.5900 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6800 - binary_accuracy: 0.6200 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6595 - binary_accuracy: 0.6200 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7340 - binary_accuracy: 0.5320 # of impressions: 2000 # of impressions: 2020 # of impressions: 2040 # of impressions: 2060 # of impressions: 2080 # of impressions: 2100 # of impressions: 2120 # of impressions: 2140 # of impressions: 2160 # of impressions: 2180 # of impressions: 2200 # of impressions: 2220 # of impressions: 2240 # of impressions: 2260 # of impressions: 2280 # of impressions: 2300 # of impressions: 2320 # of impressions: 2340 # of impressions: 2360 # of impressions: 2380 # of impressions: 2400 # of impressions: 2420 # of impressions: 2440 # of impressions: 2460 # of impressions: 2480 Updating the model at 2500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7531 - binary_accuracy: 0.5240 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6901 - binary_accuracy: 0.5520 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6713 - binary_accuracy: 0.5980 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6775 - binary_accuracy: 0.6020 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6795 - binary_accuracy: 0.5800 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6597 - binary_accuracy: 0.6020 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6707 - binary_accuracy: 0.6060 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6644 - binary_accuracy: 0.5860 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6541 - binary_accuracy: 0.6020 Epoch 10/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6560 - binary_accuracy: 0.6240 # of impressions: 2500 # of impressions: 2520 # of impressions: 2540 # of impressions: 2560 # of impressions: 2580 # of impressions: 2600 # of impressions: 2620 # of impressions: 2640 # of impressions: 2660 # of impressions: 2680 # of impressions: 2700 # of impressions: 2720 # of impressions: 2740 # of impressions: 2760 # of impressions: 2780 # of impressions: 2800 # of impressions: 2820 # of impressions: 2840 # of impressions: 2860 # of impressions: 2880 # of impressions: 2900 # of impressions: 2920 # of impressions: 2940 # of impressions: 2960 # of impressions: 2980 Updating the model at 3000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6898 - binary_accuracy: 0.5780 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6807 - binary_accuracy: 0.5960 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6743 - binary_accuracy: 0.5800 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6842 - binary_accuracy: 0.6040 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6664 - binary_accuracy: 0.6280 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6678 - binary_accuracy: 0.6260 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6553 - binary_accuracy: 0.5960 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6554 - binary_accuracy: 0.6140 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6459 - binary_accuracy: 0.6400 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6478 - binary_accuracy: 0.6180 # of impressions: 3000 # of impressions: 3020 # of impressions: 3040 # of impressions: 3060 # of impressions: 3080 # of impressions: 3100 # of impressions: 3120 # of impressions: 3140 # of impressions: 3160 # of impressions: 3180 # of impressions: 3200 # of impressions: 3220 # of impressions: 3240 # of impressions: 3260 # of impressions: 3280 # of impressions: 3300 # of impressions: 3320 # of impressions: 3340 # of impressions: 3360 # of impressions: 3380 # of impressions: 3400 # of impressions: 3420 # of impressions: 3440 # of impressions: 3460 # of impressions: 3480 Updating the model at 3500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7144 - binary_accuracy: 0.5320 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7065 - binary_accuracy: 0.5120 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7025 - binary_accuracy: 0.5760 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6843 - binary_accuracy: 0.5600 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6799 - binary_accuracy: 0.5480 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6753 - binary_accuracy: 0.5860 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6646 - binary_accuracy: 0.5980 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6763 - binary_accuracy: 0.5600 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6714 - binary_accuracy: 0.5880 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6625 - binary_accuracy: 0.5960 # of impressions: 3500 # of impressions: 3520 # of impressions: 3540 # of impressions: 3560 # of impressions: 3580 # of impressions: 3600 # of impressions: 3620 # of impressions: 3640 # of impressions: 3660 # of impressions: 3680 # of impressions: 3700 # of impressions: 3720 # of impressions: 3740 # of impressions: 3760 # of impressions: 3780 # of impressions: 3800 # of impressions: 3820 # of impressions: 3840 # of impressions: 3860 # of impressions: 3880 # of impressions: 3900 # of impressions: 3920 # of impressions: 3940 # of impressions: 3960 # of impressions: 3980 Updating the model at 4000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7030 - binary_accuracy: 0.5580 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7124 - binary_accuracy: 0.5160 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6750 - binary_accuracy: 0.5620 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6727 - binary_accuracy: 0.5720 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6641 - binary_accuracy: 0.6240 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6603 - binary_accuracy: 0.6060 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6532 - binary_accuracy: 0.6020 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6373 - binary_accuracy: 0.6100 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6415 - binary_accuracy: 0.6200 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6376 - binary_accuracy: 0.6060 # of impressions: 4000 # of impressions: 4020 # of impressions: 4040 # of impressions: 4060 # of impressions: 4080 # of impressions: 4100 # of impressions: 4120 # of impressions: 4140 # of impressions: 4160 # of impressions: 4180 # of impressions: 4200 # of impressions: 4220 # of impressions: 4240 # of impressions: 4260 # of impressions: 4280 # of impressions: 4300 # of impressions: 4320 # of impressions: 4340 # of impressions: 4360 # of impressions: 4380 # of impressions: 4400 # of impressions: 4420 # of impressions: 4440 # of impressions: 4460 # of impressions: 4480 Updating the model at 4500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7164 - binary_accuracy: 0.5260 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6955 - binary_accuracy: 0.5740 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6788 - binary_accuracy: 0.5720 Epoch 4/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6748 - binary_accuracy: 0.5860 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6714 - binary_accuracy: 0.5860 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6652 - binary_accuracy: 0.6100 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6619 - binary_accuracy: 0.6060 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6608 - binary_accuracy: 0.6120 Epoch 9/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6599 - binary_accuracy: 0.5980 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6496 - binary_accuracy: 0.6240 # of impressions: 4500 # of impressions: 4520 # of impressions: 4540 # of impressions: 4560 # of impressions: 4580 # of impressions: 4600 # of impressions: 4620 # of impressions: 4640 # of impressions: 4660 # of impressions: 4680 # of impressions: 4700 # of impressions: 4720 # of impressions: 4740 # of impressions: 4760 # of impressions: 4780 # of impressions: 4800 # of impressions: 4820 # of impressions: 4840 # of impressions: 4860 # of impressions: 4880 # of impressions: 4900 # of impressions: 4920 # of impressions: 4940 # of impressions: 4960 # of impressions: 4980 Updating the model at 5000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6688 - binary_accuracy: 0.6060 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6715 - binary_accuracy: 0.5860 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6571 - binary_accuracy: 0.6040 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6531 - binary_accuracy: 0.6360 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6430 - binary_accuracy: 0.6220 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6498 - binary_accuracy: 0.6220 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6383 - binary_accuracy: 0.6360 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6485 - binary_accuracy: 0.6140 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6403 - binary_accuracy: 0.6400 Epoch 10/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6414 - binary_accuracy: 0.6420 Trying with dropout: 0.2 # of impressions: 0 # of impressions: 20 # of impressions: 40 # of impressions: 60 # of impressions: 80 # of impressions: 100 # of impressions: 120 # of impressions: 140 # of impressions: 160 # of impressions: 180 # of impressions: 200 # of impressions: 220 # of impressions: 240 # of impressions: 260 # of impressions: 280 # of impressions: 300 # of impressions: 320 # of impressions: 340 # of impressions: 360 # of impressions: 380 # of impressions: 400 # of impressions: 420 # of impressions: 440 # of impressions: 460 # of impressions: 480 Updating the model at 500 Epoch 1/10 16/16 [==============================] - 1s 3ms/step - loss: 1.4584 - binary_accuracy: 0.5060 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.9217 - binary_accuracy: 0.4840 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7640 - binary_accuracy: 0.5220 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6963 - binary_accuracy: 0.5580 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6786 - binary_accuracy: 0.5840 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6725 - binary_accuracy: 0.6020 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6635 - binary_accuracy: 0.5920 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7161 - binary_accuracy: 0.5880 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6580 - binary_accuracy: 0.5840 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6423 - binary_accuracy: 0.6340 # of impressions: 500 # of impressions: 520 # of impressions: 540 # of impressions: 560 # of impressions: 580 # of impressions: 600 # of impressions: 620 # of impressions: 640 # of impressions: 660 # of impressions: 680 # of impressions: 700 # of impressions: 720 # of impressions: 740 # of impressions: 760 # of impressions: 780 # of impressions: 800 # of impressions: 820 # of impressions: 840 # of impressions: 860 # of impressions: 880 # of impressions: 900 # of impressions: 920 # of impressions: 940 # of impressions: 960 # of impressions: 980 Updating the model at 1000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7148 - binary_accuracy: 0.5420 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7069 - binary_accuracy: 0.5600 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7121 - binary_accuracy: 0.5400 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7156 - binary_accuracy: 0.5740 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7200 - binary_accuracy: 0.5620 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7199 - binary_accuracy: 0.5560 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6778 - binary_accuracy: 0.6040 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6632 - binary_accuracy: 0.5940 Epoch 9/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6599 - binary_accuracy: 0.6260 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6367 - binary_accuracy: 0.6200 # of impressions: 1000 # of impressions: 1020 # of impressions: 1040 # of impressions: 1060 # of impressions: 1080 # of impressions: 1100 # of impressions: 1120 # of impressions: 1140 # of impressions: 1160 # of impressions: 1180 # of impressions: 1200 # of impressions: 1220 # of impressions: 1240 # of impressions: 1260 # of impressions: 1280 # of impressions: 1300 # of impressions: 1320 # of impressions: 1340 # of impressions: 1360 # of impressions: 1380 # of impressions: 1400 # of impressions: 1420 # of impressions: 1440 # of impressions: 1460 # of impressions: 1480 Updating the model at 1500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6940 - binary_accuracy: 0.5540 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7129 - binary_accuracy: 0.5500 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6761 - binary_accuracy: 0.5860 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7081 - binary_accuracy: 0.5340 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6923 - binary_accuracy: 0.5780 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7198 - binary_accuracy: 0.5460 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7137 - binary_accuracy: 0.5680 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6967 - binary_accuracy: 0.5940 Epoch 9/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6552 - binary_accuracy: 0.6180 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6607 - binary_accuracy: 0.6340 # of impressions: 1500 # of impressions: 1520 # of impressions: 1540 # of impressions: 1560 # of impressions: 1580 # of impressions: 1600 # of impressions: 1620 # of impressions: 1640 # of impressions: 1660 # of impressions: 1680 # of impressions: 1700 # of impressions: 1720 # of impressions: 1740 # of impressions: 1760 # of impressions: 1780 # of impressions: 1800 # of impressions: 1820 # of impressions: 1840 # of impressions: 1860 # of impressions: 1880 # of impressions: 1900 # of impressions: 1920 # of impressions: 1940 # of impressions: 1960 # of impressions: 1980 Updating the model at 2000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6966 - binary_accuracy: 0.5620 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7105 - binary_accuracy: 0.5800 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7163 - binary_accuracy: 0.5920 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6696 - binary_accuracy: 0.5840 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6504 - binary_accuracy: 0.6300 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6688 - binary_accuracy: 0.6080 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6640 - binary_accuracy: 0.6340 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6558 - binary_accuracy: 0.6260 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6435 - binary_accuracy: 0.6400 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6349 - binary_accuracy: 0.6200 # of impressions: 2000 # of impressions: 2020 # of impressions: 2040 # of impressions: 2060 # of impressions: 2080 # of impressions: 2100 # of impressions: 2120 # of impressions: 2140 # of impressions: 2160 # of impressions: 2180 # of impressions: 2200 # of impressions: 2220 # of impressions: 2240 # of impressions: 2260 # of impressions: 2280 # of impressions: 2300 # of impressions: 2320 # of impressions: 2340 # of impressions: 2360 # of impressions: 2380 # of impressions: 2400 # of impressions: 2420 # of impressions: 2440 # of impressions: 2460 # of impressions: 2480 Updating the model at 2500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7073 - binary_accuracy: 0.5900 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6805 - binary_accuracy: 0.6060 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6748 - binary_accuracy: 0.6040 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6645 - binary_accuracy: 0.6060 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6674 - binary_accuracy: 0.6140 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6551 - binary_accuracy: 0.6240 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6527 - binary_accuracy: 0.6160 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6659 - binary_accuracy: 0.6040 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6660 - binary_accuracy: 0.6160 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6415 - binary_accuracy: 0.6360 # of impressions: 2500 # of impressions: 2520 # of impressions: 2540 # of impressions: 2560 # of impressions: 2580 # of impressions: 2600 # of impressions: 2620 # of impressions: 2640 # of impressions: 2660 # of impressions: 2680 # of impressions: 2700 # of impressions: 2720 # of impressions: 2740 # of impressions: 2760 # of impressions: 2780 # of impressions: 2800 # of impressions: 2820 # of impressions: 2840 # of impressions: 2860 # of impressions: 2880 # of impressions: 2900 # of impressions: 2920 # of impressions: 2940 # of impressions: 2960 # of impressions: 2980 Updating the model at 3000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6946 - binary_accuracy: 0.5920 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6908 - binary_accuracy: 0.5840 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6937 - binary_accuracy: 0.5620 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6862 - binary_accuracy: 0.5760 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6970 - binary_accuracy: 0.5740 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7177 - binary_accuracy: 0.5600 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6763 - binary_accuracy: 0.5820 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6582 - binary_accuracy: 0.6120 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6620 - binary_accuracy: 0.6040 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6574 - binary_accuracy: 0.6160 # of impressions: 3000 # of impressions: 3020 # of impressions: 3040 # of impressions: 3060 # of impressions: 3080 # of impressions: 3100 # of impressions: 3120 # of impressions: 3140 # of impressions: 3160 # of impressions: 3180 # of impressions: 3200 # of impressions: 3220 # of impressions: 3240 # of impressions: 3260 # of impressions: 3280 # of impressions: 3300 # of impressions: 3320 # of impressions: 3340 # of impressions: 3360 # of impressions: 3380 # of impressions: 3400 # of impressions: 3420 # of impressions: 3440 # of impressions: 3460 # of impressions: 3480 Updating the model at 3500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6969 - binary_accuracy: 0.5600 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6870 - binary_accuracy: 0.5700 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6798 - binary_accuracy: 0.5680 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6741 - binary_accuracy: 0.6180 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6715 - binary_accuracy: 0.5960 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6594 - binary_accuracy: 0.6280 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6649 - binary_accuracy: 0.6060 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6922 - binary_accuracy: 0.5440 Epoch 9/10 16/16 [==============================] - 0s 2ms/step - loss: 0.6498 - binary_accuracy: 0.6320 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6347 - binary_accuracy: 0.6480 # of impressions: 3500 # of impressions: 3520 # of impressions: 3540 # of impressions: 3560 # of impressions: 3580 # of impressions: 3600 # of impressions: 3620 # of impressions: 3640 # of impressions: 3660 # of impressions: 3680 # of impressions: 3700 # of impressions: 3720 # of impressions: 3740 # of impressions: 3760 # of impressions: 3780 # of impressions: 3800 # of impressions: 3820 # of impressions: 3840 # of impressions: 3860 # of impressions: 3880 # of impressions: 3900 # of impressions: 3920 # of impressions: 3940 # of impressions: 3960 # of impressions: 3980 Updating the model at 4000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6784 - binary_accuracy: 0.5860 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6692 - binary_accuracy: 0.5840 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7043 - binary_accuracy: 0.5880 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6630 - binary_accuracy: 0.5920 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6511 - binary_accuracy: 0.6280 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6511 - binary_accuracy: 0.6220 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6392 - binary_accuracy: 0.6500 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6514 - binary_accuracy: 0.5940 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6353 - binary_accuracy: 0.6320 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6330 - binary_accuracy: 0.6340 # of impressions: 4000 # of impressions: 4020 # of impressions: 4040 # of impressions: 4060 # of impressions: 4080 # of impressions: 4100 # of impressions: 4120 # of impressions: 4140 # of impressions: 4160 # of impressions: 4180 # of impressions: 4200 # of impressions: 4220 # of impressions: 4240 # of impressions: 4260 # of impressions: 4280 # of impressions: 4300 # of impressions: 4320 # of impressions: 4340 # of impressions: 4360 # of impressions: 4380 # of impressions: 4400 # of impressions: 4420 # of impressions: 4440 # of impressions: 4460 # of impressions: 4480 Updating the model at 4500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7154 - binary_accuracy: 0.5240 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7046 - binary_accuracy: 0.5360 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6845 - binary_accuracy: 0.5620 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6823 - binary_accuracy: 0.5660 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6804 - binary_accuracy: 0.5740 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6737 - binary_accuracy: 0.5640 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6827 - binary_accuracy: 0.5660 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6837 - binary_accuracy: 0.5500 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6638 - binary_accuracy: 0.5860 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6593 - binary_accuracy: 0.6000 # of impressions: 4500 # of impressions: 4520 # of impressions: 4540 # of impressions: 4560 # of impressions: 4580 # of impressions: 4600 # of impressions: 4620 # of impressions: 4640 # of impressions: 4660 # of impressions: 4680 # of impressions: 4700 # of impressions: 4720 # of impressions: 4740 # of impressions: 4760 # of impressions: 4780 # of impressions: 4800 # of impressions: 4820 # of impressions: 4840 # of impressions: 4860 # of impressions: 4880 # of impressions: 4900 # of impressions: 4920 # of impressions: 4940 # of impressions: 4960 # of impressions: 4980 Updating the model at 5000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6801 - binary_accuracy: 0.5900 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6607 - binary_accuracy: 0.6200 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6633 - binary_accuracy: 0.6220 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6600 - binary_accuracy: 0.6100 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6486 - binary_accuracy: 0.6380 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6485 - binary_accuracy: 0.6280 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6332 - binary_accuracy: 0.6400 Epoch 8/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6357 - binary_accuracy: 0.6360 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6339 - binary_accuracy: 0.6380 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6239 - binary_accuracy: 0.6560 Trying with dropout: 0.4 # of impressions: 0 # of impressions: 20 # of impressions: 40 # of impressions: 60 # of impressions: 80 # of impressions: 100 # of impressions: 120 # of impressions: 140 # of impressions: 160 # of impressions: 180 # of impressions: 200 # of impressions: 220 # of impressions: 240 # of impressions: 260 # of impressions: 280 # of impressions: 300 # of impressions: 320 # of impressions: 340 # of impressions: 360 # of impressions: 380 # of impressions: 400 # of impressions: 420 # of impressions: 440 # of impressions: 460 # of impressions: 480 Updating the model at 500 Epoch 1/10 16/16 [==============================] - 1s 3ms/step - loss: 1.2578 - binary_accuracy: 0.5080 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.8418 - binary_accuracy: 0.5180 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.8006 - binary_accuracy: 0.4920 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7543 - binary_accuracy: 0.5360 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6835 - binary_accuracy: 0.5580 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6767 - binary_accuracy: 0.6060 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6759 - binary_accuracy: 0.5760 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6739 - binary_accuracy: 0.5960 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6969 - binary_accuracy: 0.5740 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6963 - binary_accuracy: 0.6100 # of impressions: 500 # of impressions: 520 # of impressions: 540 # of impressions: 560 # of impressions: 580 # of impressions: 600 # of impressions: 620 # of impressions: 640 # of impressions: 660 # of impressions: 680 # of impressions: 700 # of impressions: 720 # of impressions: 740 # of impressions: 760 # of impressions: 780 # of impressions: 800 # of impressions: 820 # of impressions: 840 # of impressions: 860 # of impressions: 880 # of impressions: 900 # of impressions: 920 # of impressions: 940 # of impressions: 960 # of impressions: 980 Updating the model at 1000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7266 - binary_accuracy: 0.5360 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7078 - binary_accuracy: 0.5380 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7576 - binary_accuracy: 0.5300 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7017 - binary_accuracy: 0.5420 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6974 - binary_accuracy: 0.5380 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7270 - binary_accuracy: 0.5320 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7037 - binary_accuracy: 0.5580 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6529 - binary_accuracy: 0.6100 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6757 - binary_accuracy: 0.5880 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6808 - binary_accuracy: 0.5840 # of impressions: 1000 # of impressions: 1020 # of impressions: 1040 # of impressions: 1060 # of impressions: 1080 # of impressions: 1100 # of impressions: 1120 # of impressions: 1140 # of impressions: 1160 # of impressions: 1180 # of impressions: 1200 # of impressions: 1220 # of impressions: 1240 # of impressions: 1260 # of impressions: 1280 # of impressions: 1300 # of impressions: 1320 # of impressions: 1340 # of impressions: 1360 # of impressions: 1380 # of impressions: 1400 # of impressions: 1420 # of impressions: 1440 # of impressions: 1460 # of impressions: 1480 Updating the model at 1500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7177 - binary_accuracy: 0.5220 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7246 - binary_accuracy: 0.5480 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7166 - binary_accuracy: 0.5480 Epoch 4/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6939 - binary_accuracy: 0.5580 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6732 - binary_accuracy: 0.6100 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7052 - binary_accuracy: 0.5720 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6656 - binary_accuracy: 0.6000 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6968 - binary_accuracy: 0.5680 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6641 - binary_accuracy: 0.6140 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6701 - binary_accuracy: 0.5980 # of impressions: 1500 # of impressions: 1520 # of impressions: 1540 # of impressions: 1560 # of impressions: 1580 # of impressions: 1600 # of impressions: 1620 # of impressions: 1640 # of impressions: 1660 # of impressions: 1680 # of impressions: 1700 # of impressions: 1720 # of impressions: 1740 # of impressions: 1760 # of impressions: 1780 # of impressions: 1800 # of impressions: 1820 # of impressions: 1840 # of impressions: 1860 # of impressions: 1880 # of impressions: 1900 # of impressions: 1920 # of impressions: 1940 # of impressions: 1960 # of impressions: 1980 Updating the model at 2000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7108 - binary_accuracy: 0.5920 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6958 - binary_accuracy: 0.5560 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7078 - binary_accuracy: 0.5560 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6714 - binary_accuracy: 0.5880 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6547 - binary_accuracy: 0.6200 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6507 - binary_accuracy: 0.6100 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6550 - binary_accuracy: 0.6120 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6668 - binary_accuracy: 0.6420 Epoch 9/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6764 - binary_accuracy: 0.6020 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6449 - binary_accuracy: 0.6220 # of impressions: 2000 # of impressions: 2020 # of impressions: 2040 # of impressions: 2060 # of impressions: 2080 # of impressions: 2100 # of impressions: 2120 # of impressions: 2140 # of impressions: 2160 # of impressions: 2180 # of impressions: 2200 # of impressions: 2220 # of impressions: 2240 # of impressions: 2260 # of impressions: 2280 # of impressions: 2300 # of impressions: 2320 # of impressions: 2340 # of impressions: 2360 # of impressions: 2380 # of impressions: 2400 # of impressions: 2420 # of impressions: 2440 # of impressions: 2460 # of impressions: 2480 Updating the model at 2500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7030 - binary_accuracy: 0.5680 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6916 - binary_accuracy: 0.5800 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6891 - binary_accuracy: 0.5820 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6771 - binary_accuracy: 0.5980 Epoch 5/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6717 - binary_accuracy: 0.5780 Epoch 6/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6956 - binary_accuracy: 0.5660 Epoch 7/10 16/16 [==============================] - 0s 5ms/step - loss: 0.6649 - binary_accuracy: 0.6200 Epoch 8/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6590 - binary_accuracy: 0.6040 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6549 - binary_accuracy: 0.6180 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6528 - binary_accuracy: 0.6200 # of impressions: 2500 # of impressions: 2520 # of impressions: 2540 # of impressions: 2560 # of impressions: 2580 # of impressions: 2600 # of impressions: 2620 # of impressions: 2640 # of impressions: 2660 # of impressions: 2680 # of impressions: 2700 # of impressions: 2720 # of impressions: 2740 # of impressions: 2760 # of impressions: 2780 # of impressions: 2800 # of impressions: 2820 # of impressions: 2840 # of impressions: 2860 # of impressions: 2880 # of impressions: 2900 # of impressions: 2920 # of impressions: 2940 # of impressions: 2960 # of impressions: 2980 Updating the model at 3000 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6941 - binary_accuracy: 0.5740 Epoch 2/10 16/16 [==============================] - 0s 4ms/step - loss: 0.7142 - binary_accuracy: 0.5320 Epoch 3/10 16/16 [==============================] - 0s 5ms/step - loss: 0.6967 - binary_accuracy: 0.5560 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7075 - binary_accuracy: 0.5340 Epoch 5/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6652 - binary_accuracy: 0.6220 Epoch 6/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6733 - binary_accuracy: 0.5860 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6597 - binary_accuracy: 0.6160 Epoch 8/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6649 - binary_accuracy: 0.6080 Epoch 9/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6575 - binary_accuracy: 0.6100 Epoch 10/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6543 - binary_accuracy: 0.6220 # of impressions: 3000 # of impressions: 3020 # of impressions: 3040 # of impressions: 3060 # of impressions: 3080 # of impressions: 3100 # of impressions: 3120 # of impressions: 3140 # of impressions: 3160 # of impressions: 3180 # of impressions: 3200 # of impressions: 3220 # of impressions: 3240 # of impressions: 3260 # of impressions: 3280 # of impressions: 3300 # of impressions: 3320 # of impressions: 3340 # of impressions: 3360 # of impressions: 3380 # of impressions: 3400 # of impressions: 3420 # of impressions: 3440 # of impressions: 3460 # of impressions: 3480 Updating the model at 3500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7185 - binary_accuracy: 0.5280 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6977 - binary_accuracy: 0.5580 Epoch 3/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6851 - binary_accuracy: 0.5780 Epoch 4/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6784 - binary_accuracy: 0.5920 Epoch 5/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6674 - binary_accuracy: 0.6120 Epoch 6/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6563 - binary_accuracy: 0.6240 Epoch 7/10 16/16 [==============================] - 0s 5ms/step - loss: 0.6550 - binary_accuracy: 0.6260 Epoch 8/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6576 - binary_accuracy: 0.6100 Epoch 9/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6577 - binary_accuracy: 0.6100 Epoch 10/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6429 - binary_accuracy: 0.6440 # of impressions: 3500 # of impressions: 3520 # of impressions: 3540 # of impressions: 3560 # of impressions: 3580 # of impressions: 3600 # of impressions: 3620 # of impressions: 3640 # of impressions: 3660 # of impressions: 3680 # of impressions: 3700 # of impressions: 3720 # of impressions: 3740 # of impressions: 3760 # of impressions: 3780 # of impressions: 3800 # of impressions: 3820 # of impressions: 3840 # of impressions: 3860 # of impressions: 3880 # of impressions: 3900 # of impressions: 3920 # of impressions: 3940 # of impressions: 3960 # of impressions: 3980 Updating the model at 4000 Epoch 1/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6994 - binary_accuracy: 0.5740 Epoch 2/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6851 - binary_accuracy: 0.5780 Epoch 3/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6551 - binary_accuracy: 0.6000 Epoch 4/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6564 - binary_accuracy: 0.6300 Epoch 5/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6596 - binary_accuracy: 0.5980 Epoch 6/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6509 - binary_accuracy: 0.6260 Epoch 7/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6498 - binary_accuracy: 0.6120 Epoch 8/10 16/16 [==============================] - 0s 5ms/step - loss: 0.6359 - binary_accuracy: 0.6360 Epoch 9/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6306 - binary_accuracy: 0.6360 Epoch 10/10 16/16 [==============================] - 0s 5ms/step - loss: 0.6317 - binary_accuracy: 0.6500 # of impressions: 4000 # of impressions: 4020 # of impressions: 4040 # of impressions: 4060 # of impressions: 4080 # of impressions: 4100 # of impressions: 4120 # of impressions: 4140 # of impressions: 4160 # of impressions: 4180 # of impressions: 4200 # of impressions: 4220 # of impressions: 4240 # of impressions: 4260 # of impressions: 4280 # of impressions: 4300 # of impressions: 4320 # of impressions: 4340 # of impressions: 4360 # of impressions: 4380 # of impressions: 4400 # of impressions: 4420 # of impressions: 4440 # of impressions: 4460 # of impressions: 4480 Updating the model at 4500 Epoch 1/10 16/16 [==============================] - 0s 3ms/step - loss: 0.7003 - binary_accuracy: 0.5380 Epoch 2/10 16/16 [==============================] - 0s 5ms/step - loss: 0.6897 - binary_accuracy: 0.5540 Epoch 3/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6833 - binary_accuracy: 0.5660 Epoch 4/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6748 - binary_accuracy: 0.5700 Epoch 5/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6815 - binary_accuracy: 0.5760 Epoch 6/10 16/16 [==============================] - 0s 5ms/step - loss: 0.6805 - binary_accuracy: 0.5660 Epoch 7/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6892 - binary_accuracy: 0.5640 Epoch 8/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6758 - binary_accuracy: 0.5960 Epoch 9/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6561 - binary_accuracy: 0.6180 Epoch 10/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6594 - binary_accuracy: 0.6240 # of impressions: 4500 # of impressions: 4520 # of impressions: 4540 # of impressions: 4560 # of impressions: 4580 # of impressions: 4600 # of impressions: 4620 # of impressions: 4640 # of impressions: 4660 # of impressions: 4680 # of impressions: 4700 # of impressions: 4720 # of impressions: 4740 # of impressions: 4760 # of impressions: 4780 # of impressions: 4800 # of impressions: 4820 # of impressions: 4840 # of impressions: 4860 # of impressions: 4880 # of impressions: 4900 # of impressions: 4920 # of impressions: 4940 # of impressions: 4960 # of impressions: 4980 Updating the model at 5000 Epoch 1/10 16/16 [==============================] - 0s 5ms/step - loss: 0.6957 - binary_accuracy: 0.5700 Epoch 2/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6753 - binary_accuracy: 0.6080 Epoch 3/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6657 - binary_accuracy: 0.6060 Epoch 4/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6588 - binary_accuracy: 0.6260 Epoch 5/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6611 - binary_accuracy: 0.6180 Epoch 6/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6545 - binary_accuracy: 0.6160 Epoch 7/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6479 - binary_accuracy: 0.6360 Epoch 8/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6563 - binary_accuracy: 0.6120 Epoch 9/10 16/16 [==============================] - 0s 4ms/step - loss: 0.6465 - binary_accuracy: 0.6320 Epoch 10/10 16/16 [==============================] - 0s 3ms/step - loss: 0.6458 - binary_accuracy: 0.6240
In [ ]:
df_cbandits.iplot(dash = ['dash', 'solid', 'dashdot',
'dot', 'longdash', 'longdashdot'],
xTitle='Impressions',
yTitle='Cumulative Regret')
Comparison of cumulative regret with various dropout rates
The results on the above figure show that our bandit models learn after some observations how to select the ads given the user characteristics. As various dropout rates have led to different algorithm performances, an important question again becomes how to select the dropout rate. One obvious answer is to try different rates over time identify what works the best in similar online advertising problems. This approach usually works if the business has to solve similar problems again and again over a long time period. A better approach though is to learn the optimal dropout rate.