reinforcement learning
a form of machine learning that focuses on receiving feedback from the environment and updating its policy (set of actions to maximize reward)
episodic: Training is from a finite set of examples
incremental: each training example mildly changes the policy
bootstrapping: Estimate how good a state is based on how good the next state is. Don’t fully understand this, might be worth coming back to…
backup: Go from an ideal state in the future back to the current state and see what it takes to get there.
Dynamic Programming: Have a model of the world and bootstrap to figure out the optimal path to some goal state.
