Reinforcement Learning problems
MDPs, value iteration, Q-learning, policy gradients. Agents that learn by acting.
360 problems127 Easy166 Medium67 Hard
50 problems
Policy Evaluation Sweep
rl-001
Value Iteration One Sweep
rl-002
Value Iteration To Convergence
rl-003
Policy Extraction From Q
rl-004
Policy Improvement Step
rl-005
Policy Iteration One Round
rl-006
Q-Value From Transitions
rl-007
State-Value From Action Values
rl-008
Bellman Residual
rl-009
Expected Return With Discount
rl-010
Discounted Return List
rl-011
Monte Carlo Return
rl-012
TD(0) Update
rl-013
N-Step Return
rl-014
TD(Lambda) Forward View
rl-015
SARSA Update
rl-016
Q-Learning Update
rl-017
Double Q-Learning Update
rl-018
Expected SARSA Update
rl-019
Epsilon-Greedy Action
rl-020
Epsilon-Greedy Action Probabilities
rl-021
Softmax Policy Probabilities
rl-022
UCB Action Choice
rl-023
Optimistic Initialization Effect
rl-024
Bandit Sample Average Update
rl-025
Incremental Mean Update
rl-026
Reward Scaling by Maximum Absolute Value
rl-027
Advantage Function
rl-028
GAE One Step
rl-029
TD Error
rl-030
Importance Sampling Ratio
rl-031
Off-Policy Corrected Return
rl-032
Replay Buffer Sample
rl-033
Prioritized Replay Probability
rl-034
Target Network Soft Update
rl-035
DQN Loss (MSE)
rl-036
Dueling Split
rl-037
Double DQN Target
rl-038
N-Step Q Target
rl-039
REINFORCE Log-Prob Gradient
rl-040
Policy Gradient Loss
rl-041
Entropy Bonus
rl-042
Value Function Loss
rl-043
Actor-Critic Update
rl-044
A2C Advantage Normalization
rl-045
DDPG Actor Update
rl-046
TD3 Target Smoothing
rl-047
SAC Alpha Loss
rl-048
Soft Value Update
rl-049
Twin Q Min
rl-050
Showing 50 of 360 problems.