Beta Version 2.7.0

Added

  • Added “mode” parameter to the KMeans’ constructor under the “Models” section.

  • Added OneClassSupportVectorMachine under the “Models” section.

  • Added ModelParametersSafeguardWrapper under the “Others” section.

Changes

  • KMeans now accept table of matrices instead of a single matrix for its ModelParameters under the “Models” section.

  • Renamed TrainingModifier to ModelTrainingModifier under the “Others” section.

Fixes

  • Fixed some bugs and optimized ExpectationMaximization under the “Models” section.

  • Fixed a bug where these models requires an Eligibility trace object when calling episodeUpdate() and reset() functions under the “Models” section:

    • DeepQLearning

    • DeepDoubleQLearningV1

    • DeepDoubleQLearningV2

    • DeepClippedDoubleQLearning

    • DeepStateActionRewardStateAction

    • DeepDoubleStateActionRewardStateActionV1

    • DeepDoubleStateActionRewardStateActionV2

    • DeepExpectedStateActionRewardStateAction

    • DeepDoubleExpectedStateActionRewardStateActionV1

    • DeepDoubleExpectedStateActionRewardStateActionV2