#!/usr/bin/env python
# coding: utf-8

# ## Deep learning workshop

# "There's a lot more opportunity for ML than there is ML expertise in the world" (\*) - [Jeff Dean TF Summit 2018][jeff-quote]
# 
# <a href="https://youtu.be/kSa3UObNS6o?t=23m30s">
#     <img src="https://img.youtube.com/vi/kSa3UObNS6o/maxresdefault.jpg" width="400px" />
# </a>
# 
# (\*) In the context of automating model creation and replacing ML expertise with more computations!
# 
# [jeff-quote]:https://youtu.be/kSa3UObNS6o?t=23m30s

# **Goals for today**
# 
# * Experiment with deep learning in a visual way
# * Learn more about basic building blocks
# * Understand steps to build a network in PyTorch

# **What we will build today**
# 
# * Network that recognizes handwritten digits and drawings
# * Create a deep model for your own task (webcam/other set of images)

# **High-level questions**
# 
# Start with
# 
# * How do neural networks learn - building blocks
# * How can they solve such complex tasks - main ideas
# * How to build and train one in PyTorch - technical aspects
# 
# Later today
# 
# * How can they work with images - convnets
# * Can we reuse what's been learned on other tasks - transfer learning
# * What can go wrong - overfitting, interpretability

# **Workshop material**
# 
# Interactive notebooks to experiment with the material / write notes