The first assignment consisted of four parts:

1. Implementing a MLP in NumPy from the ground up by deriving forward and backward pass on paper and translating it to code
2. Implementing the same MLP with PyTorch
3. Writing a custom Layer Normalization module with manual forward and backward pass
4. Implement a VGG network architecture and compare to Transfer Learning approach

The second assignment consisted of three parts:

1. Implement Long-Short Term Networks (LSTM) as well as Bi-directional LSTM from scratch and compare their performance on a simple sequence dataset
2. Use built-in PyTorch LSTM module for text generation
3. Theoretical questions about Graph Neural Networks

The third assignment consisted of three parts:

1. Implement a Variational Auto Encoder
2. Implement a Generative Adverserial Network
3. Build a Generative Flow Based Model

Training Progress VAE on MNIST

Training Progress GAN on MNIST

Copyright © 2020 Nils Lehmann.

This project is distributed under the MIT license. In case you are a UvA student, please follow the UvA regulations governing Fraud and Plagiarism