The goal of an agent in reinforcement learning is to learn a policy that maximizes the cumulative reward it receives over time.

Q-learning, SARSA, and policy gradients are all common reinforcement learning algorithms.

Q-learning uses a value function to estimate the value of states, while SARSA uses a policy to estimate the value of state-action pairs.

The discount factor controls the trade-off between immediate and future rewards. A higher discount factor means that the agent values immediate rewards more than future rewards.

Epsilon-greedy, Boltzmann exploration, and Thompson sampling are all common exploration strategies in reinforcement learning.

Function approximation can be used to reduce the dimensionality of the state space, make the agent's policy more generalizable, and improve the agent's sample efficiency.

Linear function approximation, neural network function approximation, and kernel function approximation are all common types of function approximation used in reinforcement learning.

Model-based RL uses a model of the environment to make decisions, while model-free RL does not.

Dyna-Q is a common model-based reinforcement learning algorithm.

Q-learning is a common model-free reinforcement learning algorithm.

The actor selects actions, while the critic evaluates the value of those actions. The actor learns a policy, while the critic learns a value function. The actor is responsible for exploration, while the critic is responsible for exploitation.

DDPG, TD3, and SAC are all common types of actor-critic methods.

Intrinsic motivation can be used to encourage the agent to explore the environment, help the agent learn more efficiently, and make the agent more robust to changes in the environment.

Curiosity, progress, and competence are all common types of intrinsic motivation.

The curse of dimensionality, the exploration-exploitation trade-off, and the problem of delayed rewards are all main challenges in reinforcement learning.