Flashcards · Reinforcement Learning · Free
Reinforcement Learning flashcards, generated for you.
Example Reinforcement Learning study cards to learn from right now — then generate a full set from your own notes (plus a practice quiz) and export to Quizlet or Anki. Free, no account needed.
Example Reinforcement Learning flashcards
What is the core idea behind reinforcement learning?
An agent learns to make good decisions by interacting with an environment: it takes actions, receives rewards or penalties, and adjusts its behavior to maximize cumulative reward over time.
What is the difference between exploration and exploitation in RL?
Exploration means trying new actions to discover their effects; exploitation means choosing actions known to give high rewards. Balance is needed: too much exploration wastes time, too much exploitation misses better options.
What is a policy in reinforcement learning?
A policy is a strategy the agent follows—a mapping from states (situations) to actions. It can be deterministic (one action per state) or stochastic (probability distribution over actions).
What is the return (or cumulative reward) in RL, and why use discounting?
Return is the sum of all future rewards. Discounting (multiplying future rewards by γ < 1) makes immediate rewards more valuable than distant ones, improving learning stability and reflecting uncertainty about the future.
What is a value function, and how does it guide decision-making?
A value function V(s) estimates the expected cumulative reward from state s onward under a given policy. The agent prefers actions that lead to high-value states, turning long-term reward prediction into a local decision rule.
What is the Bellman equation, and why is it fundamental to RL?
The Bellman equation states V(s) = E[R + γV(s')], connecting a state's value to immediate reward R and the discounted value of the next state s'. It enables iterative, bootstrapping algorithms that improve estimates without waiting for episodes to finish.
What is the difference between value-based and policy-based RL methods?
Value-based methods (e.g., Q-learning) learn state or action values, then derive the policy as the action maximizing value. Policy-based methods (e.g., policy gradients) directly learn a policy parameterization without explicitly computing values.
What does a Q-function represent, and how is it updated in Q-learning?
Q(s, a) estimates the expected return from taking action a in state s. Q-learning updates it using: Q(s,a) ← Q(s,a) + α[R + γ max Q(s',a') - Q(s,a)], moving toward observed rewards and bootstrapped next-state values.
What is the advantage function, and why is it useful?
The advantage function A(s,a) = Q(s,a) - V(s) measures how much better action a is than the average action in state s. It reduces variance in learning by isolating action-specific benefits from baseline state value, improving convergence.
What problem does experience replay solve in deep RL, and how?
Experience replay stores past transitions (s, a, r, s') in a buffer and trains on random minibatches. This breaks correlation between consecutive samples, stabilizes learning from non-stationary data, and improves sample efficiency compared to online learning.
Make your own Reinforcement Learning study set
Flashcards for related topics
Studying Reinforcement Learning to build with AI? MindloomHQ turns it into real skills — structured courses, agent projects, and certificates.
Explore MindloomHQ →