Heavy-Tail Path-Adaptive Optimizer Pool / README.md

Mechanism failed

Raw ⬇ ZIP

Heavy-Tail Path-Adaptive Optimizer Pool MVP

optimizer_pool.py implements restarted diagonal AdaGrad experts with geometric horizons and the centered variance-normalized exponential meta-update from the idea.

Run:

/home/maxwelhelp/main/bin/python3 optimizer_pool.py

The script verifies two analytic predictions by sweeps:

  1. For a constant scalar gradient, cumulative AdaGrad displacement scales as sqrt(H); the asymptotic prediction is 2 alpha sqrt(H).
  2. For two fixed-loss experts with gap delta, the centered meta log-odds scale as 4 beta sqrt(T) under the stated per-expert variance accumulators.

It also reports a deliberately falsifiable dynamic-recovery sweep and a shifted quadratic comparison. The latter is exploratory only; the pool uses multiple expert updates and is not a claim of equal-FLOP superiority.