Session 2: Neural Networks for Classification

Lesson 2

The second lesson introduces implements neural networks, building from the simplest 1-layer linear network in the previous lesson up to deep recurrent networks with non-linear activations. We will interact with real neurophysiological data throughout.

Lesson 2.1: Putting together a deep network

* Other activation functions
    * Logistic
    * RELU
* Calculating gradient
* Stacking layers
* BatchNorm
* Dropout
* ...

Lesson 2.2: Convolutional Neural Network

CNN

• Using single-channel ECoG and 1-D (temporal) convolutions.
• Using multi-channel ECoG with spatio-temporal convolutions.

Lesson 2.3: Recurrent Neural Network

• GRU
• LSTM