About this Project

• When I was teaching myself machine learning, I wanted to make sure that I fully understood the logic and backpropagation calculus behind all of the network architectures and training techniques that I implemented.
• This project uses pure Python and NumPy to implement Graph Convolutional Networks with global and hierarchical architectures as well as various graph pooling methods.
• The models were trained to classify compounds using the Adadelta algorithm for gradient descent.