Players and Adversaries

An AI that is capable of learning from its experiences, starting with basic knowledge and can then adapt.

Power: Anything providing an advantage, positive.

Costs: Anything providing a disadvantage, negative.
(Shadow Costs: Hidden disadvantages)

A Feedback loop that causes a system to change further in the same direction

The training of machine learning models to make a sequence of decisions. Tests how desirable certain "plans" are (the likelihood of performing a given action when a player with a certain skill is in a certain state)

An extension of DTMC.

A graph representing the decision-making model where decisions (or actions) are taken to navigate a sequence of states under the assumption that the property holds.
In reinforcement learning, these transitions have a name and an associated reward.

As the number of state variables in the system increases, the size of the system state space grows exponentially, leading to a state explosion.

We can avoid/Reduce this by minimising the AI agent, removing unreachable, dead, and undistinguishable states

Returns a two element array which contains the minimum and the maximum, two players trying to achieve one of them.

A method for eliminating branches during the search that will not affect the outcome; Minimax with Alpha-Beta pruning always returns the same scores as Minimax would when using the same depth limit; Potentially much faster than minimax.

Alpha and Beta are best possible outcome for Max and Min in each other's turns. Any value that is higher (or lower) than that floor are pruned when this is replaced.

A variation of the minimax algorithm, for use in artificial intelligence systems that play two-player zero-sum games, such as backgammon, in which the outcome depends on a combination of the player's skill and chance elements such as dice rolls.

For multi-player games with more than 2 people.

For each final configuration, the score associated to each player n in an n dimensional score vector.

Set the world as a "passive player" which performs no action but opponent actions affect the score whenever one of the players changes its own score.