This type of histogram automatically adapts bins when new values are added. Note that only fixed-width continuous binning scheme is currently supported.
# Necessary import evil
import physt
from physt import h1, h2, histogramdd
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline
# Create an empty histogram
h = h1(None, "fixed_width", bin_width=10, name="People height", axis_name="cm", adaptive=True)
h
Histogram1D(bins=(0,), total=0.0, dtype=None)
# Add a first value
h.fill(157)
h.plot()
h
Histogram1D(bins=(1,), total=1.0, dtype=None)
# Add a second value
h.fill(173)
h.plot();
# Add a few more values, including weights
h.fill(173, 2)
h.fill(186, 5)
h.fill(188, 3)
h.fill(193, 1)
h.plot(errors=True, show_stats=True);
ha = h1(None, "fixed_width", bin_width=10, adaptive=True)
ha.plot(show_stats=True);
# Beginning
ha.fill_n([10, 11, 34])
ha.plot();
# Add a distant value
ha.fill_n([234], weights=[10])
ha.plot(show_stats=True);
# Let's create a huge dataset
values = np.random.normal(130, 20, 100000)
%%time
# Add lots of values (no loop in Python)
hn = h1(None, "fixed_width", bin_width=10, adaptive=True)
hn.fill_n(values)
# ha.plot()
CPU times: user 12.8 ms, sys: 3.62 ms, total: 16.4 ms Wall time: 15.3 ms
%%time
# Comparison with Python loop
hp = h1(None, "fixed_width", bin_width=10, adaptive=True)
for value in values:
hp.fill(value)
CPU times: user 4.23 s, sys: 50.1 ms, total: 4.28 s Wall time: 4.29 s
# Hopefully equal results
print("Equal?", hp == hn)
hp.plot(show_stats=True);
Equal? True
ha1 = h1(None, "fixed_width", 5, adaptive=True)
ha1.fill_n(np.random.normal(100, 10, 1000))
ha2 = h1(None, "fixed_width", 5, adaptive=True)
ha2.fill_n(np.random.normal(70, 10, 500))
ha = ha1 + ha2
fig, ax= plt.subplots()
ha1.plot(alpha=0.1, ax=ax, label="1", color="red")
ha2.plot(alpha=0.1, ax=ax, label="2")
ha.plot("scatter", label="sum", ax=ax, errors=True)
ax.legend(loc=2); # TODO? Why don't we show the sum???