Intelligent agents
Intelligent Agents¶
Key points:
- Agent and agent function (percept -> action mapping)
- Performance measure evaluates the behavior of the agent in the environment. Maximised by rational agent.
- Task environment. It can be:
- Fully or partially observable
- Deterministic or stochastic
- Single or multi agent
- Episodic or sequential
- Static or dynamic
- Discrete or continuous
- Known or unknown
- Agent types:
- Simple reflex based
- Model-based
- Goal-based, eg. planner
- Utility-based, eg. minimax
- Learning agents -> improve performance measure via 4 components:
- Performance element
- Critic (give feedback)
- Learning element (add new rules to world)
- Problem generator (explore domain)
- State representations:
- Atomic -> search, game-playing, hidden Markov models, MDP
- Factored -> CSP, planning, propositional logic, Bayesian networks, ML
- Structured -> relational databases, first-order logic, first-order probability models, knowledge-based learning, NL