Hello,
It depends on your problem. What does the 50 vertical and 432 horizontal matrices you mentioned mean? If your matrix is an image you should use a two-dimensional CNN but probably it is not. I'm just assuming to give you an idea that the matrix is a time-series data (for example, an estimate of temperature) probably 432 represents the time axis of this data, and 50 is the factors affecting this temperature (pressure, altitude, etc), in this case, the input matrix would be the vertical matrix containing 50 elements.
