Class | Description |
---|---|
DistortRandomizer |
This class provides distort randomization technique, which distorts existing
weight values using specified distortion factor.
|
GaussianRandomizer |
This class provides Gaussian randomization technique using Box Muller method.
|
HeZhangRenSunUniformWeightsRandomizer |
Sources:
https://arxiv.org/abs/1502.01852 Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification
https://stats.stackexchange.com/questions/47590/what-are-good-initial-weights-in-a-neural-network
https://github.com/keras-team/keras/blob/master/keras/initializers.py
|
NguyenWidrowRandomizer |
This class provides NguyenWidrow randmization technique, which gives very good results
for Multi Layer Perceptrons trained with back propagation family of learning rules.
|
RangeRandomizer |
This class provides ranged weights randomizer, which randomize weights in specified [min, max] range.
|
WeightsRandomizer |
Basic weights randomizer, iterates and randomizes all connection weights in network.
|
Copyright © 2019 Neuroph Project. All rights reserved.