Is there any layer defined in matlab for sine activation function? Or else can we define the layer using functionLayer?

How we can use a layer with sine activation function? Is it possible to use the functionLayer to define a sine activation function or should I define a class for creating the layer as shown in https://in.mathworks.com/help/deeplearning/ug/define-custom-deep-learning-intermediate-layers.html?

Answer 1

[Matt J]

If it will not have any learnable parameters, you can use a functionLayer.

[BIPIN SAMUEL]

Thank You @Matt J

I be created a customized layer based on attention mechanism for deep learning application using https://in.mathworks.com/help/deeplearning/ug/define-custom-deep-learning-layer.html. But while checking the validity of layer using the following code:

layer=CoAtten(Name="atten");
validInputSize = [1 14 1024];
layout = networkDataLayout(validInputSize,"CBT");
layer = initialize(layer,layout);
checkLayer(layer,validInputSize,ObservationDimension=3)

The last line (checkLayer) shows the error like this- ""The value of 'Layer' is invalid. Layers that require formatted dlarray inputs are not supported."

What this error shows? Is the custom layer is invalid? I have used stripdims inside the predict fuction while creating the layer. Is it happening due to this?

[Matt J]

What about,

checkLayer(layer,layout,ObservationDimension=3)

[Matt J]

And why ObservationDimension=3 when you have a Batch dimension that is the second dimension?

[BIPIN SAMUEL]

Thank You @Matt J, I have used checkLayer from https://in.mathworks.com/help/deeplearning/ref/checklayer.html for the checking the validity of the layer. I just want to know whether the layer that I have created is correct or not. I have given ObservationDimension=3 to show the size of dlarray input but I am not sure whether it is correct or not.

While using that command "The value of 'Layer' is invalid. Layers that require formatted dlarray inputs are not supported." - this error is showing, but I am not sure that what it represents whether the error is with the layer or with the way I have used the checkLayer(layer,layout,ObservationDimension=3) command. Is there any other way to check the customized layer is working or not?

[Matt J]

There are too many details hidden from us to know what's going on.

[BIPIN SAMUEL]

ok @Matt J, This is the custom layer that I have created:

classdef CoAtten < nnet.layer.Layer  ...
        & nnet.layer.Formattable ... 
        %   & nnet.layer.Acceleratable 
    % properties
    %     % (Optional) Layer properties.
    % 
    %     % Declare layer properties here.
    % end
    properties (Learnable)
        % (Optional) Layer learnable parameters.
        alpha
    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 = CoAtten(NameValueArgs)
            % (Optional) Create a myLayer.
            % This function must have the same name as the class.
            arguments
                NameValueArgs.Name = '';
            end
            name=NameValueArgs.Name;
            layer.Name=name;
            layer.Description="Attention Mechanism based on correlation";
            layer.Type="Correlation Attention";
        end
        function layer = initialize(layer,layout)
            % (Optional) Initialize layer learnable and state parameters.
            if isempty(layer.alpha)
                idx=finddim(layout,"C");
                numChannels = layout.Size(idx);
                layer.alpha=dlarray(zeros(numChannels));
            end
        end
        
        function Z = predict(layer,varargin)
            % Forward input data through the layer at prediction time and
            % output the result and updated state.
            %
            X=varargin;
            filtersize=ones(1);
            nc=ones(1);
            numfilters=ones(1);
            sz=[filtersize,nc,numfilters];
            numout=prod(filtersize)*numfilters;
            numin=prod(filtersize)*numfilters;
            weights=initializeGlorot(sz,numout,numin);
            bias=dlarray(zeros(nc));
            query=dlconv(X,weights,bias,WeightsFormat='CUT');
            weights=initializeGlorot(sz,numout,numin);
            key=dlconv(X,weights,bias,WeightsFormat='CUT');
            weights=initializeGlorot(sz,numout,numin);
            value=dlconv(X,weights,bias,WeightsFormat='CUT');
            X=stripdims(X);
            query=stripdims(query);
            key=stripdims(key);
            value=stripdims(value);
            X_zscore=zscore(X);
            X_transpose=permute(X_zscore,[1 3 2]);
            X_flatten=permute(X_transpose,[2 3 1]);
            query_flatten=permute(query,[2 4 1 3]);
            query_energy=pagemtimes(X_flatten,query_flatten);
            query_energy=permute(query_energy, [3 1 4 2]);
            key_flatten=permute(key,[2 4 1 3]);
            key_energy=pagemtimes(X_flatten,key_flatten);
            key_energy=permute(key_energy,[3 1 4 2]);
            query_energy=permute(query_energy,[2 3 1]);
            key_energy=permute(key_energy,[2 4 1 3]);
            energy=pagemtimes(query_energy,key_energy);
            energy=permute(energy,[3 1 4 2]);
            energy=permute(energy, [2 1 3]);
            attention=softmax(energy,'DataFormat',"UCU");
            value_flatten=permute(value,[2 3 1]);
            out=pagemtimes(value_flatten,attention);
            out=permute(out,[2 1 3]);
            Alpha=layer.alpha;
            Z=Alpha*out+X;
            Z=dlarray(Z,"CBT");
        end
    end
end

This is the layer that I have used for the validity check using "checkLayer".

[Matt J]

Does this make any more sense?

layer=CoAtten(Name="atten");
validInputSize = [1 14 1024];
layout = networkDataLayout(validInputSize,"CBT");
layer = initialize(layer,layout);
checkLayer(layer,layout,ObservationDimension=2)

Skipping GPU tests. No compatible GPU device found.
 
Skipping code generation compatibility tests. To check validity of the layer for code generation, specify the CheckCodegenCompatibility and ObservationDimension options.
 
Running nnet.checklayer.TestLayerWithoutBackward
..
================================================================================
nnet.checklayer.TestLayerWithoutBackward/formattableLayerPredictIsFormatted(Observations=one) was filtered.
    Test Diagnostic: Test did not run because 'predict' threw an error.
================================================================================
.
================================================================================
nnet.checklayer.TestLayerWithoutBackward/formattableLayerPredictIsFormatted(Observations=multiple) was filtered.
    Test Diagnostic: Test did not run because 'predict' threw an error.
================================================================================
....
================================================================================
Verification failed in nnet.checklayer.TestLayerWithoutBackward/predictDoesNotError(Observations=one).
    ----------------
    Test Diagnostic:
    ----------------
    Test failure may be due to the layer not being initialized. If the layer is not initialized, then initialize it by calling its initialize method.
    ---------------------
    Framework Diagnostic:
    ---------------------
    The function 'predict' threw an error:
    Undefined function 'initializeGlorot' for input arguments of type 'double'.
    
    Error in CoAtten/predict (line 57)
                weights=initializeGlorot(sz,numout,numin);
    ------------------
    Stack Information:
    ------------------
    In /MATLAB/toolbox/nnet/cnn/+nnet/+checklayer/TestLayerWithoutBackward.m (TestLayerWithoutBackward.predictDoesNotError) at 26
================================================================================
.
================================================================================
Verification failed in nnet.checklayer.TestLayerWithoutBackward/predictDoesNotError(Observations=multiple).
    ----------------
    Test Diagnostic:
    ----------------
    Test failure may be due to the layer not being initialized. If the layer is not initialized, then initialize it by calling its initialize method.
    ---------------------
    Framework Diagnostic:
    ---------------------
    The function 'predict' threw an error:
    Undefined function 'initializeGlorot' for input arguments of type 'double'.
    
    Error in CoAtten/predict (line 57)
                weights=initializeGlorot(sz,numout,numin);
    ------------------
    Stack Information:
    ------------------
    In /MATLAB/toolbox/nnet/cnn/+nnet/+checklayer/TestLayerWithoutBackward.m (TestLayerWithoutBackward.predictDoesNotError) at 26
================================================================================
.. ....
================================================================================
nnet.checklayer.TestLayerWithoutBackward/predictIsConsistentInType(Precision=single,Device=cpu) was filtered.
    Test Diagnostic: Test did not run because 'predict' threw an error.
================================================================================
.
================================================================================
nnet.checklayer.TestLayerWithoutBackward/predictIsConsistentInType(Precision=double,Device=cpu) was filtered.
    Test Diagnostic: Test did not run because 'predict' threw an error.
================================================================================
..... ..
================================================================================
nnet.checklayer.TestLayerWithoutBackward/backwardPropagationDoesNotError(Observations=one) was filtered.
    Test Diagnostic: Test did not run because 'predict' threw an error.
================================================================================
.
================================================================================
nnet.checklayer.TestLayerWithoutBackward/backwardPropagationDoesNotError(Observations=multiple) was filtered.
    Test Diagnostic: Test did not run because 'predict' threw an error.
================================================================================
.
Done nnet.checklayer.TestLayerWithoutBackward
__________

Failure Summary:

     Name                                                                                                Failed  Incomplete  Reason(s)
    =================================================================================================================================================
     nnet.checklayer.TestLayerWithoutBackward/formattableLayerPredictIsFormatted(Observations=one)                   X       Filtered by assumption.
    -------------------------------------------------------------------------------------------------------------------------------------------------
     nnet.checklayer.TestLayerWithoutBackward/formattableLayerPredictIsFormatted(Observations=multiple)              X       Filtered by assumption.
    -------------------------------------------------------------------------------------------------------------------------------------------------
     nnet.checklayer.TestLayerWithoutBackward/predictDoesNotError(Observations=one)                        X                 Failed by verification.
    -------------------------------------------------------------------------------------------------------------------------------------------------
     nnet.checklayer.TestLayerWithoutBackward/predictDoesNotError(Observations=multiple)                   X                 Failed by verification.
    -------------------------------------------------------------------------------------------------------------------------------------------------
     nnet.checklayer.TestLayerWithoutBackward/predictIsConsistentInType(Precision=single,Device=cpu)                 X       Filtered by assumption.
    -------------------------------------------------------------------------------------------------------------------------------------------------
     nnet.checklayer.TestLayerWithoutBackward/predictIsConsistentInType(Precision=double,Device=cpu)                 X       Filtered by assumption.
    -------------------------------------------------------------------------------------------------------------------------------------------------
     nnet.checklayer.TestLayerWithoutBackward/backwardPropagationDoesNotError(Observations=one)                      X       Filtered by assumption.
    -------------------------------------------------------------------------------------------------------------------------------------------------
     nnet.checklayer.TestLayerWithoutBackward/backwardPropagationDoesNotError(Observations=multiple)                 X       Filtered by assumption.
Test Summary:
	 16 Passed, 2 Failed, 6 Incomplete, 10 Skipped.
	 Time elapsed: 1.3986 seconds.

[BIPIN SAMUEL]

Rectified Thank You very much @Matt J