Release Version 2.13
Models
-
Added these new options for “sigmoidFunction” in LogisticRegression’s new() constructor:
-
HardSigmoid
-
SoftSign
-
Swish
-
ArcTangent
-
Swish
-
BipolarSigmoid
-
-
All models now initialize ModelParameters when calling predict() function.
-
NearestCentroid model now returns cost array when calling train() function.
-
All of the models’ divergence warning is now only printed when “isPrintOutput” parameter is true in BaseModel’s new() constructor.
-
All models now print nan (not a number) value warning when “isPrintOutput” parameter is true in BaseModel’s new() constructor.
-
Optimized KMedoids by reducing the amount of calculation needed to calculate the cost.
Cores
- Added DistanceFunctionDictionary and ZTableFunction.
Quick Setups
- Fixed a bug where calling reset() function from ReinforcementLearningBaseQuickSetup gives an error.
All
- The error messages now screams at you less.