matlab DRSN:how to cutom my layer?

Has anyone ever built a DRSN with matlab？I have some trouble，I do not know how to customize the need for soft threshold network？I always get an error on my custom layer.

classdef softthresholdingLayer < nnet.layer.Layer % ...

% & nnet.layer.Formattable ... % (Optional)

% & nnet.layer.Acceleratable % (Optional)

properties

% (Optional) Layer properties.

% Declare layer properties here.

end

properties (Learnable)

% (Optional) Layer learnable parameters.

% Declare learnable parameters here.

% thresholding

end

properties (State)

% (Optional) Layer state parameters.

% Declare state parameters here.

end

properties (Learnable, State)

% (Optional) Nested dlnetwork objects with both learnable

% parameters and state parameters.

% Declare nested networks with learnable and state parameters here.

end

methods

function layer = softthresholdingLayer(numInputs,name)

% (Optional) Create a myLayer.

% This function must have the same name as the class.

% Define layer constructor function here.

layer.NumInputs = numInputs;

layer.Name = name;

% layer.Description = "soft thresholding";

% Initialize layer weights.

% layer.Weights = rand(1,1);

end

function Z = predict(layer,vargin)

X = vargin;

% Initialize output

X1 = X{1};

th = X{2};

% x = extractdata(gather(x));

% th = extractdata(gather(th));

Z = 1.*(X1>th) + 0.*(X1<=th & X1>=(-th)) + 1.*(X1<-th);

% Z = dlarray(Z,'SSCB');

end

end

end

softthresholdingLayer/predict：Too many input parameters

Is that because I can't use “predict”？

Answer 1

[Venu]

Hi @BB

The "predict" method in your custom layer should only take two arguments: layer and X. Your implementation attempts to take an unspecified number of input arguments with "vargin", which seems to be a typo for MATLAB's variable input argument "varargin". This is not the correct approach for layer methods in MATLAB's Deep Learning Toolbox. Instead, the input to the layer should be directly passed as the second argument.

function Z = predict(layer, X)

Hope this helps!