Notes on selected aspects of ongoing work

Welcome to the notes section. Click on any note below to view it as a webpage or download it as a PDF.

A Hands-On Reproduction of Width–Depth Separation in ReLU Networks Learning Iterated Maps

We track the epochal evolution of the decision boundaries of a 3-3-1 MLP, pre-trained for one task, as it learns a new task.

We discuss elementary aspects of representing logic gates with perceptrons