Coxeter Artin Research Fellowship

Motivation

Our research lab set out to discover if Transformer's or other Machine Learning Architectures were able to learn a pattern for reducing words in different Coxeter groups.

Approach

TBD

Features

Models Made: RNN, LSTM, and Transformer Models built for the Task

Developed a sophisticated word generation algorithm for Non-Trivial Coxeter groups of length $n$

Used the Transform Lens Mechanistic Interpretability Library to analyze the attention matrices of trained models.

Links

-EXPMMLL (BC Math Lab)

• Project Site
• Writeup
• Repository