lecture 1 + 2 Flashcards
What do we need to define for agents to perform a certain task? What do they map?
A performance measure, a set of sensors, actuators and it will operate in an environment. A simple agent maps the percepts its given via sensors to actions.
Why do we need to evaluate an agent function? What is important with this?
This allows us to see how an agent performs over time easily. It is important we pick a performance measure which the agent can not somehow “cheat” at in order to get a high score without doing what we wanted.
What is a reflex agent, state based agent, reasoning agent, goal based agent, and linguistic agent?
Reflex: link sensors to actuators
State based: link sensors to internal variables(known as state) which links to actuators
Reasoning agents: Can store and manipulate representations of situations in the world.
Goal based: Can search for a way to achieve desired situation
Linguistic agents: Can exchange representations with one another.
Why can perception and action be blurred?
Actions change the environment, this consequently changes what the agent perceives.
What is an emergent behaviour?
A behaviour which results from the interaction between the agent and environment.
What is the subsumption architecture?
Lots of seperate reflex agent functions stacked on top of each other, each of which is known as a behaviour.
What are the main factors environments vary on?
Fully observable vs partially deterministic vs stochastic episodic vs sequential offline vs online discrete vs continuous single-agent vs multi-agent
Why is it important to work in the real world and not simulations?
Hard to simulate all relevant aspects of environment, unexpected problems and unexpected benefits.
