AI Agents flashcards, generated for you.

A program that perceives its environment, makes decisions, and takes actions to reach a goal—like a robot that sees obstacles and navigates around them.

A cycle where the agent observes the current state, decides on an action, executes it, receives feedback/reward, and repeats—the fundamental loop all agents follow.

Reactive agents respond immediately to current observations (if-then rules); deliberative agents plan ahead by modeling future states and consequences before acting.

A rule or function that maps states to actions—it tells the agent what to do in each situation. Denoted π(a|s), it gives the probability of choosing action a in state s.

A function V(s) that estimates the long-term reward an agent will receive from state s onward. Agents use it to evaluate which states are worth being in and guide decision-making toward better futures.

The insight that a state's value equals its immediate reward plus the discounted value of the next state. Mathematically: V(s) = R(s) + γ·V(s'). It breaks the problem of long-term planning into immediate + future components.

The dilemma agents face: exploit (use known good actions) vs. explore (try new actions to learn better ones). Balancing both is key—pure exploitation misses better strategies; pure exploration wastes time.

Model-based: agent learns a model of the environment (how actions lead to next states) and plans using it. Model-free: agent learns a policy or value function directly from experience without building an explicit model.

A mathematical framework with states, actions, transition probabilities, and rewards. An MDP assumes the Markov property: the next state depends only on the current state and action, not history. Most agent formalisms use MDPs as their foundation.

On-policy: agent learns the value of the policy it is currently following (e.g., SARSA). Off-policy: agent learns the value of an optimal policy while following a different exploratory policy (e.g., Q-learning). Off-policy is more sample-efficient but harder to stabilize.