Custom Machine Learning Algorithm + Curated ML Knowledge Base
A repository combining a custom experimental machine learning algorithm for tabular data with a hand-curated, machine-readable knowledge base of roughly 900 university courses, research papers and explainers, each normalized into Markdown with full provenance.
This project has two connected layers. Adaptive Prototype Residual Learning, or APRL, is a custom classifier and regressor for numeric tabular data with local nonlinear residuals, uncertainty estimates, handling of missing values and incremental batch updates, backed by eleven tests and a reproducible five-dataset benchmark against baselines.
The second layer is a curated knowledge base of 923 documents, about 11 million tokens in total, spanning leading university courses, canonical research papers and widely cited explainers, normalized into consistent Markdown with clean metadata and full provenance on every file. It includes 391 arXiv papers, 474 lecture transcripts across 14 courses and channels, and dozens of web articles, tagged into a 17-topic map that can be opened as an Obsidian vault or pointed at an AI agent for grounded answers.