Skip to content

Instantly share code, notes, and snippets.

View shivvor2's full-sized avatar

LJZ shivvor2

View GitHub Profile
@shivvor2
shivvor2 / drug_extraction.ipynb
Created November 15, 2025 23:31
Jupyter notebook demoing a workflow for struct extraction from FDA's authoritative drug database
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
@shivvor2
shivvor2 / algorithmic_equivalence.md
Created November 13, 2025 17:20
Algorithmic equivalence of our implementation of DenMune and the algorithm presented in the original paper

Algorithmic Equivalence and Optimized Implementation

This document outlines how our new implementation of DenMune is a faithful, high-performance translation of the conceptual algorithm presented in the original paper. A direct, line-by-line implementation of the paper's pseudocode would be algorithmically naïve and unacceptably slow for practical use.

Our implementation replaces inefficient loops with vectorized operations and modern data structures while preserving the core logic.

1. Foundational Graph Construction

The initial steps of the algorithm involve building a k-nearest neighbor graph and deriving properties from it. Our implementation replaces the paper's implied loops with optimized scikit-learn and scipy equivalents.

@shivvor2
shivvor2 / denmune_implementation_issues.md
Last active December 19, 2025 03:22
Implementation issues for scikit-learn-contrib/denmune-clustering-algorithm

Necessity of a complete reimplementation of the DenMune algorithm

The current implementation of scikit-learn-contrib/denmune-clustering-algorithm is a extremely problematic. It violates multiple sklearn design principles, is algorithmically inefficient, and is coded in an unmaintainable way.

A complete rewrite is required to adhere to scikit-learn-contrib standards

1. Violation of the sklearn standards

1.1. No inheritance from BaseEstimator/ClusterMixin

@shivvor2
shivvor2 / main.ipynb
Created May 8, 2025 07:17
A colab notebook containing code for RL finetuning Qwen-2.5-0.5B-Instruct for long reasoning capabilities (both full finetuning and Lora Fientuning)
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.