Module fr.inria.MochyUi
Package fr.inria.mochy.statsAndTasks
-
Class Summary Class Description AbstractStats FullNnetPipelineTask generate 1000 random neural nets calculate for each the value for 100 runs : time elapsed + beta*|target speed - average speed| + beta*|target speed - average speed after equalization| select the 5 ones with the minimal values optimize them 3 times by updating their weights if it leads to a best value keep the best one of the 5 optimized neural netsGenerateNeuralTask generate random neural networks, only those wich make the equalization network to get to an equalization are saved in the neural folder.GenerateNnetMultiTokensNbTask MultipleRunsTask MultipleTokensNbTask OptimizeNeuralNetTask add/remove 0.1 to the weights and save the new network if it has better results called from OptimizeNeuralController and MultipleRunsController.equalizationFullStatsOptimizeNnetMultiTokensNbTask StatsTask TimeElapsedAlphaMultiTokens