domingo, dezembro 13, 2015

Lidar com a incerteza (parte I)

"When an agent chooses its strategy from a fixed set of possible ones, its decision has to take several factors into account: (i) locality, i.e. the subset of those agents, the respective agent most likely interacts with, (ii) heterogeneity, i.e. the possible strategies of these agents, (iii) time horizon, i.e. the number of foreseen interactions with these agents. The best strategy for a rational agent would be the one with the highest payoff for a given situation. To reach this goal, however, can be di fficult for two reasons: uncertainty, i.e. particular realizations are usually subject to random disturbances and can therefore be known only with a certain probability, and co-evolution, i.e. the agent does not operate in a stationary environment, but in an ever-changing one, where its own action influences the decisions of other agents and vice-versa.
To reduce the risk of making the wrong decision, it often seems to be appropriate just to copy the successful strategies of others. Such an imitation behavior is widely found in biology, but also in cultural evolution."
Trecho retirado de "Distribution of Strategies in a Spatial MultiAgent
Game" de Frank Schweitzer e Robert Mach.

Sem comentários: