rl-blogs 14
- Why TD Learning Likes the 2-Norm: Mean Directions, Inner Products, and Bellman Geometry
- Linear TD vs Neural TD: A Tale of Two Geometries
- From Snail Trails to Robust POMDPs: Safe Learning with Hidden Monsters
- TD Learning Is Almost Gradient Descent: A Finite-Time View of Linear TD
- Bellman Operators and Bellman Optimality
- Discounted vs. Average Reward Reinforcement Learning
- Concentration Inequalities: A Researcher's Guide from Markov to Freedman
- The One-Pixel Attack: Fooling a Neural Network by Changing One Pixel
- Why Do Neural TD Converge ?
- Function Approximation in RL: From Tables to Linear Models to Neural Networks
- The Beauty of a Simple Proof: TD Learning Without Projection
- Stochastic Gradient Descent: Why Randomness Works
- Why Gradient Descent Works: A Small Mathematical Story
- Why Vanilla Q-Learning Breaks Under Corrupted Rewards