Beta Version 1.15.0
Added
-
Added experience replay capabilities to SARSA neural network and Q-learning neural network models.
-
Added OneVsAll under “Others” to convert any binary classification models into multi class ones.
-
Added ExpectedStateActionRewardStateActionNeuralNetwork model.
Fixes
-
Fixed KMedoids and KMeans calculations.
-
Fixed bugs related to Affinity Propagation model.
Changes
-
Changed the ModelChecking to ModelChecker. Also changed a lot of its functionalities.
-
Changed the GradientDescentModes to GradientDescentModifier. Also changed a lot of its functionalities.
-
Some of the classification models now returns predictedLabelVector as opposed to single value.
-
Made the codes for clustering models more readable.
-
Changed the outputs for a number of classification and clustering algorithms.
-
Removed SupportVectorMachineOneVsAll and LogisticRegressionOneVsAll.
-
Changed the way how DBSCAN and AffinityPropagation store its model parameters.