Package io.github.kirstenali.deepj.loss

Class MSELoss

java.lang.Object

io.github.kirstenali.deepj.loss.MSELoss

All Implemented Interfaces:

LossFunction

public class MSELoss
extends Object
implements LossFunction

Mean-Squared-Error loss.

Reduction convention: the loss is averaged over every element (rows × cols), not just over rows. Consequently the gradient is 2·(predicted − actual) / (rows·cols). This keeps the loss scale independent of the feature dimension. If a sum reduction or a per-row mean is required, scale the result accordingly.

Constructor Summary

Constructors

Constructor	Description
MSELoss()

Method Summary

Modifier and Type	Method	Description
Tensor	gradient(Tensor predicted, Tensor actual)
float	loss(Tensor predicted, Tensor actual)

Methods inherited from class java.lang.Object

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

Constructor Details

MSELoss

public MSELoss()

Method Details

loss

public float loss(Tensor predicted,
 Tensor actual)

Specified by:

loss in interface LossFunction

gradient

public Tensor gradient(Tensor predicted,
 Tensor actual)

Specified by:

gradient in interface LossFunction