Exercises
Here are some Deep learning exercises in PyTorch. We recommend trying them after you have gone through the Introduction to Machine Learning lectures this week on Neural Networks - it will make better sense then!
The exercises are split into three parts.
We recommend you to try at least Part 1, where you construct a Multilayer Preceptron for a simple 2D classification task.
Part 2 is more advanced, where you train a CNN classifier to classify real images. Will be useful if you would like to work on Computer Vision problems.
In Part 3, you will construct an Autoencoder. Unlike Parts 1 and 2, this is a form of unsupervised learning. The Introduction to Machine Learning course does not cover this particular topic (this can be considered a Dimensionality Reduction problem).
The exercises are in the form of a Jupyter notebook hosted on a Google Colab. Click here to access the notebook. You can make a copy of the notebook to your own Google Drive, and run it on Google Colab yourselves.
A big thanks to Harry Coppock and Luca Grillotti for preparing these exercises!