How to concatenate features from one fullyConnectedLayer in a DNN with inputs being images from one class and features from the second class for classifier training?training?

%temp2.m

imageInputSize = [28,28,1];

filterSize = 3;

numFilters = 8;

numClasses = 10;

numFeatures = 50;

layers = [

imageInputLayer(imageInputSize,'Normalization','none','Name','images')

convolution2dLayer(filterSize,numFilters,'Name','conv')

reluLayer('Name','relu')

fullyConnectedLayer(50,'Name','fc1')%1 x 1 x 50 x N

squeezeLayer()%50 x N

concatenationLayer(2,2,'Name','cat')

fullyConnectedLayer(numClasses,'Name','fc2')

softmaxLayer('Name','softmax')

classificationLayer];

lgraph = layerGraph(layers);

featInput = featureInputLayer(numFeatures,Name="features");%3 x N

lgraph = addLayers(lgraph,featInput);

lgraph = connectLayers(lgraph,"features","cat/in2");

numObservations = 100;

fakeImages = randn([imageInputSize,numObservations]);%28 28 1 100

imagesDS = arrayDatastore(fakeImages,IterationDimension=4);

fakeFeatures = randn([numFeatures,numObservations]);%100 x 50

featureDS = arrayDatastore(fakeFeatures,IterationDimension=2);%50x100

fakeTargets = categorical(mod(1:2*numObservations,numClasses));%1x100

targetDS = arrayDatastore(fakeTargets,IterationDimension=2);

ds = combine(imagesDS,featureDS,targetDS);

opts = trainingOptions("adam","MaxEpochs",1,"MiniBatchSize",128);

net=trainNetwork(ds,lgraph,opts);

function layer = squeezeLayer(args)

arguments

args.Name='';

end

layer = functionLayer(@squeezeLayerFcn,"Name",args.Name,"Formattable",true);

end

function x = squeezeLayerFcn(x)

x = squeeze(x);

% Since squeeze will squeeze out some dimensions, we need to relabel x.

% Assumption: x does not have a 'T' dimension.

n = ndims(x);

newdims = [repelem('S',n-2),'CB'];

x = dlarray(x,newdims);

%dims(x)

end

Error in temp2 (line 35)

net=trainNetwork(ds,lgraph,opts);

Caused by:

Layer 'cat': Input size mismatch. Size of input to this layer is different from the expected input size.

Inputs to this layer:

from layer 'layer' (size 50(C) × 1(B))

from layer 'features' (size 50(C) × 1(B))

Answer 1

[Ranjeet]

Hi Ming,

Assuming that you want to concatenate ‘features’ and output of ‘squeezeLayer’, changing the 1st argument of concatenationLayer((2,2,'Name','cat') to concatenationLayer((1,2,'Name','cat') should solve the issue regarding input size mismatch.

Moreover, you can use analyzeNetwork that does the network analysis and present any error by plotting the network. For your case, you can use analyzeNetwork(lgraph).