2025.07.04 meeting notes

Last week

• $T$ sampled from random uniform, like hurst $T\sim U(100,550)$
• prioritized experience replay
• instead of 100, ..., 550 for $T$ do $2^k$
• each model iferred on all other $T$‘s
• finish rewrite in C
• sweep hyperparams in C

Notes

• checkers
• asteroids
• hardmaze
• double cart pole (wip)

TODO

• try to evaluate the optimal strategy on C env
• write interfacing python program to get fbm relaization from c to python