Artificial neural networks#

In this session we learn about artificial neural networks.

Learning goals for this session#

  1. Become familiar with ANNs:

    1. mathematical notation in matrix-vector form

    2. weights & biases (slopes & intercepts), score, activation function, hidden layers, prediction

  2. Be able to use PyTorch to implement a feed-forward ANN:

    1. building the model by hand

    2. using built-in helper functions (nn.Module, DataLoader …)

This unit requires basic familiarity with concepts and notation from linear algebra. To recap, there is a short section on algebra recap with a handout and some further references.

Slides#

Here are the slides for this session.

Practical exercises#

There are two notebooks for exercises. First, we will implement a multi-layer feed-forward network “by hand”. Then, we will implement the same model (for the same training data) by using PyTorch’s helper functions.