01
Chris Nota · 2018 · ML Review
Why?
Start here: what bias and variance mean in RL, and how GAE, value-function bootstrapping, and baselines identify and mitigate them.
Loading…
One moment while we fetch this page.
Theory and math for the bias-variance tradeoff in (deep) reinforcement learning: definitions, high-variance policy gradients, and n-step bootstrapping.
For
Beginners who know the RL basics and want a short path into bias, variance, GAE, baselines, and bootstrapping.
Covers
Bias and variance in RL, variance reduction for policy gradients, and n-step / bootstrap tradeoffs.
The full collection in the community’s reading order.
01
Chris Nota · 2018 · ML Review
Why?
Start here: what bias and variance mean in RL, and how GAE, value-function bootstrapping, and baselines identify and mitigate them.
02
Balaji AI · 2021 · balajiai.github.io
Why?
Rigorous math proofs deriving baselines and GAE from scratch.
03
Ryan Lee · 2021 · ryanlee.ai
Why?
Closes the loop on n-step bootstrapping: how the return horizon trades bias for variance in practice.