Package ai.djl.training.initializer

Class TruncatedNormalInitializer

java.lang.Object

ai.djl.training.initializer.TruncatedNormalInitializer

All Implemented Interfaces:

Initializer

public class TruncatedNormalInitializer
extends java.lang.Object
implements Initializer

Naive implementation of a truncated normal initializer. Simply samples from a normal distribution and throws away anything outside two standard deviations.

See Also:

https://en.wikipedia.org/wiki/Truncated_normal_distribution

Field Summary

Fields inherited from interface ai.djl.training.initializer.Initializer

ONES, ZEROS

Constructor Summary

Constructors

Constructor	Description
TruncatedNormalInitializer()	Creates an instance of TruncatedNormalInitializer with a default sigma of 0.01.
TruncatedNormalInitializer(float sigma)	Creates a TruncatedNormalInitializer initializer.

Method Summary

Modifier and Type	Method	Description
NDArray	initialize(NDManager baseManager, Shape shape, DataType dataType)	Initializes a single NDArray.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

TruncatedNormalInitializer

public TruncatedNormalInitializer()

Creates an instance of TruncatedNormalInitializer with a default sigma of 0.01.

TruncatedNormalInitializer

public TruncatedNormalInitializer(float sigma)

Creates a TruncatedNormalInitializer initializer.

Parameters:

sigma - the standard deviation of the truncated normal distribution. Values outside (-2σ, 2σ) will be rejected.

Method Detail

initialize

public NDArray initialize(NDManager baseManager,
                          Shape shape,
                          DataType dataType)

Initializes a single NDArray.

Specified by:

initialize in interface Initializer

Parameters:

baseManager - the NDManager to create the new NDArray in

shape - the Shape for the new NDArray

dataType - the DataType for the new NDArray

Returns:

the NDArray initialized with the manager and shape