Software
These public repositories connect theoretical results to reproducible experiments, training pipelines, and evaluation tools. Each project is paired with the corresponding research record.
Python
Reference implementation for masked, uniform-replacement, and Gaussian noise paths, including Markov, text8, and CIFAR-10 experiments and tools for reproducing the paper’s conservation-law evaluations.
Python
Training and evaluation code for structured neural polar decoders, with mutual-information estimation and input-distribution or code-rate optimization for channels with and without memory.
Python
Experiments for deletion, insertion-deletion-substitution, and multi-read DNA storage channels, with learned decoding built around polar-code structure.
See my GitHub profile for additional research repositories and ongoing open-source work.