What means MinLength in sequenceInputLayer?

Environment

• MATLAB R2023b
• Deep Learning Toolbox

Question

I try to use this example program to learn complex-valued neural networks.

https://jp.mathworks.com/help/deeplearning/ug/train-network-with-complex-valued-data.html

At line 38, layers are defined as below and the argments of `sequenceInputLayer` are

• numChannels
• SplitComplexInputs
• MinLength

layers = [ ...
    sequenceInputLayer(numChannels,SplitComplexInputs=true,MinLength=minLength)
    convolution1dLayer(filterSize,numFilters,Padding="causal")
    reluLayer
    layerNormalizationLayer
    convolution1dLayer(filterSize,2*numFilters,Padding="causal")
    reluLayer
    layerNormalizationLayer
    globalAveragePooling1dLayer
    fullyConnectedLayer(numResponses)
    regressionLayer];

What means `MinLength` in `sequenceInputLayer`?

I tried `MinLength=1` but this program worked.

If MinLength meant MinLength for down sampling, `MinLength=1` hasn't must work.

Answer 1

[Pratyush Swain]

Hi mika,

The "MinLength" argument is used to define the smallest length of input sequences that the network can process. When you set "MinLength=1", it means that the network will accept sequences that are at least one time step long. It shows the capability of the network to handle variable-length input sequences.

The "MinLength" argument's default value is 1 itself.When you train or assemble a network, it checks that sequences of length 1 can propagate through the network.

For more information on "sequenceInputLayer" and its arguments, please refer to https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layer.sequenceinputlayer.html