Skip to content

Deep Learning🔗

In this class, we construct simple deep ANNs using the PyTorch library on the Fashion MNIST example.

Notebook [source] [Colab]

Neural Architecture Slides

Instead of calculating backpropagation by hand as in the first class, this class uses automatic differentiation built in to PyTorch. This is built on the autograd package which has its own tutorial.

One important concept is convolutional neural network layers. The Stanford CS231N class has a good interactive demonstration of convolution. This page shows demonstrations of stride, padding, and dilation.

To facilitate the use of Torch, we introduce ignite at the end of class. An example of using ignite is given in this notebook

Additional Resources🔗

The deep learning book is fully available online and contains many great examples. Notebook versions of those examples are available here. Chapter 9 specifically covers convolutional neural networks.