Very simple version of the code used for the experiments in the paper Hidden Activations Are Not Enough: A General Approach to Neural Networks Predictions.

Drafts of textbooks. One on the math inspired by supersymmetric field theories. One on low dimensional (below 3+1) integrable physics.

Search for non-trivial elements of the kernel of the Burau representation of the four-strand braid group.

AIDN is a deep learning algorithm to represent any finitely-presented algebraic object with a set of deep neural networks.

A Python-based Rubik's Cube simulator and visualizer using matrix algebra for efficient representation, manipulation, and visualization of the cube's state. Includes a group theory approach to cube operations and a proof-of-concept visualization module using Matplotlib.

Linear-time sequence modeling that replaces attention's O(n²d) complexity with O(nd) summation-based aggregation. Demonstrates constraint-driven emergence: how functional representations can develop from optimization pressure and architectural constraints alone, without explicit pairwise interactions.

many powerful tools for studying irreducible representations of SU(n), including making animations of hadron flavor-state multiplets

Computes representation matrices for Lie groups

Compute the knowledge matrices associated to a neural network.

Isomorphisms of quiver representations applied to neural networks.

Tools for exploiting Morphological Symmetries in robotics

On-the-fly generator of space-group irreducible representations