Models for 2d(N x M) matrix inputs and (N x 1) outputs

I have many (N x M) matrix as features and (N x 1) matrix as the lables.
I want to train a model, the input is (N x M) matrix, and the model can predict the (N x 1) matrix.
Assume each row contains features of an item, so the matrix contains N items, the labels are the probabilities for each item.

Any advices? How to implement this model?