EasyDeepLearn
Lesson

Reinforcement Learning

214 questions12 sections

MDPs, exploration, policy vs value methods, and modern actor-critic algorithms.

Every questionFlashcardsTake the quiz

Introduction

Reinforcement learning studies agents that learn from interaction: given a state, pick an action, receive a reward, repeat. The framework — a Markov Decision Process — is deceptively small and hides real subtlety around exploration, credit assignment, and off-policy correction.

Modern successes (AlphaGo, RLHF for language models, robotics) all rely on the ideas below. Interview questions typically test whether you understand the *tradeoffs*: on-policy vs off-policy, model-based vs model-free, value vs policy, greedy vs stochastic.

The 12 sections

Each section is a short read on one subject, with every answer written out. Work through them in order, or jump to the one you are weakest on.

  1. 01Fundamentals & MDPsDefine the RL setting: agent, environment, state, action, reward, policy.56 questions3 easy17 medium36 hard
  2. 02Value methods (Q-learning, DQN)What update does Q-learning perform?39 questions1 easy8 medium30 hard
  3. 03Policy gradients & actor-criticWhat is a policy gradient and why do we use it?25 questions13 medium12 hard
  4. 04Exploration strategiesWhat is the exploration-exploitation tradeoff?11 questions2 easy5 medium4 hard
  5. 05Model-based & planningMonte Carlo Tree Search — how it works in AlphaGo/MuZero.4 questions2 medium2 hard
  6. 06Reward design & shapingWhat is reward shaping and what are its pitfalls?3 questions2 medium1 hard
  7. 07Deep RL engineeringLearning-rate schedule for PPO / SAC — what works?7 questions3 medium4 hard
  8. 08Offline RLWhat is off-policy evaluation and why is it hard?22 questions4 medium18 hard
  9. 09Multi-agent RLMulti-agent RL — what makes it fundamentally different?7 questions7 hard
  10. 10RLHF & LLM alignmentWhat is RLHF and how does it fit into training an LLM?24 questions5 medium19 hard
  11. 11Applications & safetyAlphaGo — what made it different from prior Go engines?11 questions5 medium6 hard
  12. 12Interview scenariosInterview: 'design an RL system for X' — what's your framework?5 questions4 medium1 hard