Definition
This concept originates from reinforcement learning and involves an agent interacting with an unknown environment. The automaton selects actions from a finite set and receives a penalty or reward signal. Based on this feedback, it adjusts the probability distribution over its actions using a learning algorithm, gradually converging toward the optimal action that yields the highest expected reward. It serves as a foundational block for more complex multi-agent systems.
Summary
A learning automaton is a simple stochastic decision-making unit that iteratively updates its action probabilities based on environmental feedback to maximize reward.
Key Concepts
- Action probability vector
- Reward/penalty signal
- Stochastic optimization
Use Cases
- Resource allocation problems
- Network routing optimization
- Simple control systems