rands

(To be removed) Symmetric random weight/bias initialization function

rands will be removed in a future release. For more information, see Transition Legacy Neural Network Code to dlnetwork Workflows.

For advice on updating your code, see Version History.

Syntax

W = rands(S,PR)
M = rands(S,R)
v = rands(S)


Description

rands is a weight/bias initialization function.

W = rands(S,PR) takes

S	Number of neurons
PR	R-by-2 matrix of R input ranges

and returns an S-by-R weight matrix of random values between –1 and 1.

M = rands(S,R) returns an S-by-R matrix of random values. v = rands(S) returns an S-by-1 vector of random values.

Examples

Here, three sets of random values are generated with rands.

rands(4,[0 1; -2 2])
rands(4)
rands(2,3)

Network Use

To prepare the weights and the bias of layer i of a custom network to be initialized with rands,

1. Set net.initFcn to 'initlay'. (net.initParam automatically becomes initlay’s default parameters.)

2. Set net.layers{i}.initFcn to 'initwb'.

3. Set each net.inputWeights{i,j}.initFcn to 'rands'.

4. Set each net.layerWeights{i,j}.initFcn to 'rands'.

5. Set each net.biases{i}.initFcn to 'rands'.

To initialize the network, call init.

Version History

Introduced before R2006a

R2026a: rands will be removed

rands will be removed in a future release.

To learn more about initializing and customizing neural network learnable parameters, see these examples and topics:

• Compare Layer Weight Initializers

• Specify Custom Weight Initialization Function

• Neural Network Weight Tying

• Initialize Learnable Parameters for Model Function

In most cases, the default initializers for layer learnable parameters work well. To initialize learnable parameters to be random values between –1 and 1, define this function.

function weights = initializeSymmetric(sz)

weights = 2*rand(sz) - 1;

end

To use this initialization function in a deep learning layer, use the corresponding initializer property of the layer. For example, to use this function to initialize the weights in a convolution2dLayer object, set the WeightsInitializer argument to @initializeSymmetric.

For more information, see Transition Legacy Neural Network Code to dlnetwork Workflows.

Version History

Introduced before R2006a

R2026a: rands will be removed

rands will be removed in a future release.

To learn more about initializing and customizing neural network learnable parameters, see these examples and topics:

• Compare Layer Weight Initializers

• Specify Custom Weight Initialization Function

• Neural Network Weight Tying

• Initialize Learnable Parameters for Model Function

For more information, see Transition Legacy Neural Network Code to dlnetwork Workflows.