Post-training 0/0

Post-training

RL foundations

Markov decision processes, value functions and the Bellman equations, then value iteration, policy iteration, Q-learning and SARSA. Policy gradients and RLHF are built on these definitions.

0 / 12 solved

Questions

0/12 solved
Question Module Tags Difficulty Success
1. The Dice Roller MDPs & Value Functions
mdptransitions
Easy —
2. Five Now Or Twenty Later MDPs & Value Functions
reinforcement-learningmdp +2
Medium —
3. Expected Reward Under A Stochastic Policy MDPs & Value Functions
reinforcement-learningmdp +2
Medium —
4. Averaging Q Into V MDPs & Value Functions
reinforcement-learningbellman +2
Medium 50%
5. Seven Over Four Dynamic Programming
policy-evaluationbellman-expectation +2
Medium —
6. One Sweep Then Greedy Dynamic Programming
policy-iterationbellman-equation +2
Medium 40%
7. Two Sweeps Of Value Iteration Dynamic Programming
reinforcement-learningdynamic-programming +2
Medium —
8. One Sample Is Not An Expectation Dynamic Programming
reinforcement-learningdynamic-programming +2
Medium —
9. One Episode, Two Updates Model-Free Methods
reinforcement-learningmonte-carlo +2
Medium —
10. Point Nine Two Five Model-Free Methods
epsilon-greedyexploration +2
Medium —
11. Same Transition, Two Updates Model-Free Methods
reinforcement-learningq-learning +2
Medium —
12. Cliff Notes Model-Free Methods
q-learningsarsa +2
Medium —