AASDL Program

November 5th: Ian Goodfellow

Morning, 9:00-11:00 Intro to Deep Learning

• Maximum likelihood
• Stochastic gradient descent
• Deep architectures
• Backpropagation
• LSTMs

Morning, 11:30-14:00: Computer Vision

• Convolution
• Pooling
• Maxout
• Network-in-network
• Inception
• Batch normalization
• Image captioning

Afternoon, 15:00-17:00: Intro to Theano, computer vision exercises

•  Work through the Theano exercises: https://github.com/goodfeli/theano_exercises
• Implement and train a small convolutional network to classify CIFAR-10

November 6th: Navdeep Jaitly

Morning, 9:00-11:00

• The hybrid model for speech recognition 
• DNN  based acoustic models
• LSTM based acoustic models
• Recent developments in end to end speech recognition

Afternoon, 11:30-14:00

• Building and training a basic DNN based acoustic model 
• Building a recognizer in Theano

Afternoon, 15:00-17:00: Speech recognition exercises