Neuronal Network - Time reslved Multiple Input (mxn), Targets only a random points in time

Hi everybody and thanks for reading!

My challange is to model wear of objects, like drills. I'm not sure if it is solvable with a neuronal network.

I have time resolved input data which the wear is correlating with. And I have targets which are the measurements of the wear after some time. So I do not have a target for each time-step, but lots of rows of inputs.

e.g.:

I have n=20 input tables, always 3 columns and variating amount of rows. For each input table I have one target value. Is there a way to create a predictive model for that using NN? The current solution is to create an alytical model with which I calculate a wear for each row, sumarize the wear for each input table and then optimize with the global optimization. However, I'm not happy with the results.

Thanks and best wishes Michael

[Greg Heath]

Please explain your problem in terms of the MATLAB convention of N pairs of I-dimensional "I"nputs and corresponding O-dimensional "O"utput targets.

 [ I N ] = size(input)

 [ O N ] = size(target)

Thanks

Greg