Independent researcher working on the Neural Percolation Model (NPM) — a framework that transfers percolation-theory concepts from porous media science to the analysis of neural network weight structure and emergence.
Background: M.S. in Engineering Thermophysics, Dalian University of Technology. Senior engineer working at the intersection of simulation and AI.
Research Program · NPM
The NPM research program investigates how neural network weight distributions form a skeleton that governs emergent capabilities. Key observations include a cross-family narrow band in the Weibull shape parameter k observed across 10+ pretrained models.
Detailed technical work is ongoing. See publications below.