python - Can correlation matrix be used as an input to Convolutional Neural Network (CNN)

Question

I have 10000 2D matrices, belonging to three different classes, of size n * m. For every matrix the number of columns (m) is constant but the number of rows (n) varies to a large range. As I have a very limited computational capacity, so is it possible to compute the pairwise correlation between each column, thus reducing the matrix to size m * m, and input these correlation matrices to convolutional neural networks for multiclass classification. My prime objective is to use the differences between columns, of different classes, for classification.

question from:https://stackoverflow.com/questions/65645636/can-correlation-matrix-be-used-as-an-input-to-convolutional-neural-network-cnn

Answer 1

Waitting for answers