| Name | Description |
|---|---|
| BasicRandomizer | Provides basic functionality that most randomizers will need. |
| ConsistentRandomizer | |
| ConstRandomizer | A randomizer that will create always set the random number to a const value, used mainly for testing. |
| Distort | A randomizer that distorts what is already present in the neural network. |
| GaussianRandomizer | Generally, you will not want to use this randomizer as a pure neural network randomizer. More on this later in the description. Generate random numbers that fall within a Gaussian curve. The mean represents the center of the curve, and the standard deviation helps determine the length of the curve on each side. This randomizer is used mainly for special cases where I want to generate random numbers in a Gaussian range. For a pure neural network initializer, it leaves much to be desired. However, it can make for a decent randomizer. Usually, the Nguyen Widrow randomizer performs better. Uses the "Box Muller" method. http://en.wikipedia.org/wiki/Box%E2%80%93Muller_transform Ported from C++ version provided by Everett F. Carter Jr., 1994 |
| NguyenWidrowRandomizer | Implementation of Nguyen-Widrow weight initialization. This is the default weight initialization used by Encog, as it generally provides the most trainable neural network. |
| RangeRandomizer |