Creating CNN architecture for binary classification

I’m reaching out to kindly ask if somone could review the CNN architecture I’ve implemented in MATLAB. The code is running as expected, but I’d appreciate your expert opinion to confirm whether the structure is sound and appropriate for the task.

Below is a snippet of the architecture and training configuration:

matlab

CopyEdit

%% === CNN Architecture ===

layers = [

sequenceInputLayer(1, 'Name', 'input', 'MinLength', minTrainLen)

convolution1dLayer(5, 32, 'Padding', 'same', 'Name', 'conv1')

batchNormalizationLayer('Name', 'bn1')

reluLayer('Name', 'relu1')

maxPooling1dLayer(2, 'Stride', 2, 'Name', 'pool1')

convolution1dLayer(3, 64, 'Padding', 'same', 'Name', 'conv2')

batchNormalizationLayer('Name', 'bn2')

reluLayer('Name', 'relu2')

dropoutLayer(0.3, 'Name', 'dropout1')

globalAveragePooling1dLayer('Name', 'gap')

fullyConnectedLayer(32, 'Name', 'fc1')

reluLayer('Name', 'relu3')

fullyConnectedLayer(2, 'Name', 'fc_output')

softmaxLayer('Name', 'softmax')

classificationLayer('Name', 'output')

];

%% === Training Options ===

options = trainingOptions('adam', ...

'InitialLearnRate', 1e-3, ...

'MaxEpochs', 30, ...

'MiniBatchSize', max(1, min(64, numel(XTrainFinal))), ...

'Shuffle', 'every-epoch', ...

'ValidationData', {XVal, YVal}, ...

'ValidationFrequency', 5, ...

'ValidationPatience', 2, ...

'Verbose', false, ...

'Plots', 'none', ...

'ExecutionEnvironment', 'auto');

[Matt J]

I believe you already asked that here,

https://www.mathworks.com/matlabcentral/answers/2178755-is-my-cnn-correct?s_tid=srchtitle

The statement of the question hasn't changed much, so I don't think you will get very different answers.