Suppose that a region rectangle is a matrix of size n_lat * n_lon (pixels). I have the out put of a climate model for the region for k days. so my data is of size n_lat * n_lon * k.
I have also the observations for the k days as matrices n_lat * n_lon. Without vectorizing the matrices how can I use CNN to fit to data?
I need a code based on Keras and Tensorflow for constructing a CNN in order to get data in matrix format not columnar and train it. This is a type of regression problem not classification.