I am trying to apply 1d-cnn to predict the output neutron flux passing through different shielding thicknesses. I have bunch of input spectra (1d vector of normalized weights at different energies) passing through a shielding and create a flux (also a 1d vector of energy (numerical values at different energies)). My 1d-cnn was able to train with different input weights at different energies and their corresponding output weights (fluxes at different energies). Now, I have data for the input and output weights (fluxes) at different thicknesses and want to predict if I pass a set of weights at a given thickness (which is not one of the trained thicknesses) to predict the output flux. Basically an interpolation with 1d-cnn that can understand the relationship between the attenuation capabilities of different thicknesses of shielding and predict the output at any arbitrary thickness.

The first level of the problem was done with 1d-cnn network when my only variation is fluxes of input and output as a function of energy. Now, I want to add another dimension (which is basically a label) and interpolate the results between different labels.