01_Introduction_to_Deep_Learning.md

1. Introduction to Deep Learning

[TOC]

Why deep learning

• Input output shape is flexible
• Better performance when data is enough

Outline of this class

• Backpropagation
• Computer vision
• Recommender Systems
• NLP
• Optimization: Theory, methods and tricks
• Generative models and unsupervised learning

Framworks and computation graphs

Some frameworks

• Automatic differentiation : TensorFlow, MXnet, CNTK, Theano
• Dynamic and high level: Torch & PyTorch, Chainer, MinPy, DyNet..
• Keras: high level frontend for 4 automatic differentiation Framworks

Compoutation Graph