change.txt

24/04/2018

====== Version 0.3 =======

*********************************
Performance Improvements
time execution reduced by 75%

Internal Architecture Vectorized

EJML Updated 0.30 -> 0.34
*********************************

Neural Network
- removed outputAll

Feedforward
    * outputLayers -> activations
    + activationsDropout
    + fixed getParameters

Layers
    + outputDropout

Initializations
    + HeInit

LossFunctions
    + Improvement CrossEntropy
    + Added Logistic Loss Function
    + MSE vectorized
    + RMSE vectorized

Backpropagation
    + removed gradient
    + removed minibatch
    + fixed L2 regularization
    + Performance incresed

Training Algorithms
    + Conjugate Gradient Added
    + Scaled Conjugate Gradient Added
    + Scaled Conjugate Gradient with minibatch added
    - Clean SGD
    + SGD Numeric Added
    - TrainingALgorithm Reorganized

Beta Rules Added for Conjugate Gradient
	+Fletcher Reeves
	+PolakRebiere

Gradients
	+ DropoutGradient
	+ NoiseGradient
	+ NumericGradient
	+ StandardGradient
	+ Gradient
	+ GradientFactory

Update Rules
	+ Momentum
	+ SGC
	+ Update
	+ UpdateRule
	
TransferFunctions
	update EJML 0.30 -> 0.34
	
Test Added