ribs.schedulers

Schedulers compose the entire QD algorithm together.

Specifically, the scheduler performs two roles. First, the scheduler facilitates the interaction between the archive and the population of emitters. The scheduler adds solutions generated by emitters to the archive and passes the results of evaluation and archive addition to the emitters. Second, schedulers can select which emitters generate new solutions on each iteration of the algorithm. For instance, schedulers can make decisions on active emitters based on how well each emitter performs in previous iterations.

Scheduler(archive, emitters[, ...])

A basic class that composes an archive with multiple emitters.

BanditScheduler(archive, emitter_pool[, ...])

Schedules emitters with a bandit algorithm.

BayesianOptimizationScheduler(archive, emitters)

Similar to Scheduler but with support for upscaling archives.