MATLAB CNN Classification Different Results When Shuffled Columns

Hello everyone,

I have applied 1D CNN on speech recognition, when I shuffled columns I got diffrent results, for example, using matrix(:,[1 2 3]) gives different classification results than matrix(:,[2 3 1]) which should be the same, I have tried rng("default") but it hasn't resolved the issue. Can someone please assist me with this?

Thank you in advance!

[Hamza]

@Pujitha Narra

[Hamza]

@MathWorks Support Team

Answer 1

[Avadhoot]

Hi Hamza,

I understand that you have tried to use 1D CNN with shuffled columns and got different results in each case.

A CNN uses filters to extract features from the input data. The filters are applied to different regions of the input, and the output is a feature map that represents the activation of the filters. The order of the columns in the input data affects how the filters are applied and how the output feature map is constructed.

Consider this example: if you pass an image to a CNN model, it will classify it as class A. However, if you pass the same image with flipped coordinates for each pixel, the CNN will most likely classify it as a different class because it leverages local features through its filters.

To obtain consistent results, it is recommended to use other techniques like RNNs or LSTMs for speech recognition.

I hope it helps.

[Hamza]

@Avadhoot Hi, thanks for your answer, I have tried LSTM and I faced the same issue!!

[Hamza]

@MathWorks Support Team