reinforcement learning编辑历史

Reinforcement learning (RL) is a subfield of machine learning and artificial intelligence in which an agent learns to make sequences of decisions by interacting with an environment. The agent receives rewards or penalties for its actions and aims to maximize cumulative reward over time. Unlike supervised learning, RL does not rely on labeled input-output pairs; instead, it explores its environment, discovers which actions yield the most reward, and refines its policy through trial and error. RL has been successfully applied to domains such as game playing, robotics, autonomous navigation, and recommendation systems.

## 1 Core concepts

### 1.1 Agent and environment
In reinforcement learning, the **agent** is the learner or decision-maker that interacts with the **environment**, which comprises everything outside the agent. The agent observes the environment's state, chooses actions, and the environment responds with a new state and a reward signal.

### 1.2 State and action
A **state** is a representation of the environment at a given time step. An **action** is a move or decision that the agent can take. The set of all possible states is the state space, and the set of all possible actions is the action space. States may be fully or partially observable.

### 1.3 Reward and return
A **reward** is a scalar feedback signal received after each action, indicating the immediate benefit or cost of that action. The **return** is the cumulative discounted sum of rewards over time, typically defined as \(G_t = \sum_{k=0}^{\infty} \gamma^k R_{t+k+1}\), where \(\gamma\) (0 ≤ γ ≤ 1) is the discount factor that balances immediate versus future rewards.

### 1.4 Policy
A **policy** (\(\pi\)) is a strategy that the agent uses to decide which action to take in a given state. It can be deterministic (\(\pi(s) = a\)) or stochastic (\(\pi(a|s)\) = probability of taking action \(a\) in state \(s\)). The goal of RL is to find an optimal policy that maximizes expected return.

###