SoftMax: SoftMax

In frhl/neuralnetr: Neural Nets in R

Description

Classic softmax that can be interpreted as a probability distribution over n items

For 'SoftMax$class_fun()', given the column vector of class probabilities for each point (computed by Softmax), return a vector of the classes (integers) with the highest probability for each point.

Super class

neuralnetr::ClassModule -> SoftMax

Methods

Public methods

• SoftMax$forward()

• SoftMax$backward()

• SoftMax$class_fun()

• SoftMax$clone()

Inherited methods

• neuralnetr::ClassModule$sgd_step()

---

Method forward()

Usage

SoftMax$forward(Z)

Arguments

Z

a vector of pre-activations.

Returns

probabilities of the given class

---

Method backward()

Usage

SoftMax$backward(dLdZ)

Arguments

dLdZ

a vector of gradients dLdZ.

Returns

dLdZ

---

Method class_fun()

a classify function.

Usage

SoftMax$class_fun(Ypred)

Arguments

Ypred

a vector of predictions.

Returns

the index for the vector of the predicted label.

---

Method clone()

The objects of this class are cloneable with this method.

Usage

SoftMax$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

See Also

Other activation: ReLU, Sigmoid, Tanh

frhl/neuralnetr documentation built on Nov. 9, 2020, 2:24 p.m.