The earlier research focused on preference neuron networks with a fixed number of modules, and networks evolved using one form of Module Mutation. This paper builds on earlier results showing that modular neural networks can be successfully evolved for Ms. The number of preference neuron modules can be fixed, or discovered using Module Mutation (also called Mode Mutation ). when to use which module) can be based on a human-specified task division similar to that used in Multitask Learning, or discovered automatically through the use of special neurons that indicate the network’s preference for using each module. Each module represents a different policy, and the agent can use one at a time. In contrast, this paper evolves neural networks with multiple output modules using a framework called Modular Multiobjective NEAT (MM-NEAT). Although it is possible to represent multimodal behavior with such policies, it is difficult to do so. Pac-Man regardless of whether ghosts are threatening or edible. Despite the need for multi-modal behavior, most learning approaches to the game have focused on learning monolithic policies that control Ms. In other words, multiple distinct modes of behavior are required. The switch in game dynamics requires a switch in play strategy. Pac-Man is usually the prey of the ghosts, but if she eats a power pill, the situation is reversed: Ghosts temporarily become her prey. Pac-Man is a predator-prey scenario, with a twist. Pac-Man is interesting because simple rules give rise to a game in which complex strategies are needed to succeed. This popularity extends to AI research, as evidenced by numerous papers and two different competitions. Pac-Man is among the most popular video games of all time. The results demonstrate that MM-NEAT can discover interesting and effective behavior for agents in challenging games.
Interestingly, the best networks dedicate modules to critical behaviors (such as escaping when surrounded after luring ghosts near a power pill) that do not follow the customary division of the game into chasing edible and escaping threat ghosts. Both fixed modular networks and Module Mutation networks outperform monolithic networks and Multitask networks. Several versions of Module Mutation are evaluated in this paper. The appropriate number of modules can be fixed or discovered using a genetic operator called Module Mutation. Multitask) or discovered automatically by evolution. The modules are used at different times according to a policy that can be human-designed (i.e. In contrast, this paper uses a framework called Modular Multi-objective NEAT (MM-NEAT) to evolve modular neural networks. Pac-Man have treated the game as a single task to be learned using monolithic policy representations. Past approaches to learning behavior in Ms. Pac-Man must escape ghosts when they are threats and catch them when they are edible, in addition to eating all pills in each level. Pac-Man is a challenging video game in which multiple modes of behavior are required: Ms.
0 kommentar(er)
