Shake AI Lab

Hands-on experiments in machine learning, geometry, and dynamical systems. Each post measures something in code and ships an interactive visualization. Originally written in Japanese; selected pieces are translated here.

Jul 31, 2026What a Tesseract's Shadow Loses: Flattening a 4D Hypercube to 2D, PCA Lost to All 200 Random ProjectionsQuantifying what a 4D→3D→2D projection of a tesseract throws away: perspective-projection edge distortion (65%–150%), a stress comparison of four dimensionality-reduction methods, how degradation grows with dimension, and why only 4D rotations that involve w look like stretching.