This repo documents presentable experiments in the Reinforcement Learning training of players under the scope of https://github.com/honglu2875/hironaka

Highlights

Blowup trees of some checkpoints of AlphaZero agents

Contents

first batch of a prototype v0

Small scale, against choosefirst v0

Small scale with varying batch sizes

Scaling up small nets

AlphaZero style RL (notes).

The blowup tree of E8 surface singularity using Zeillinger policy