The deep learning model does this by first multiplying the columns of input with a weight and then adding them together (if you have three columns three groups of weights are defined). Then there are activations. Then there are other layers. Finally, a layer maps the hidden values to two different groups that you have as the output. If your measure is MSE, then MSE of all the values (between the observed and predicted values in two outputs) is computed and used as the function of optimization.

It is a very rough description and a lot of details are skipped.

Hello, thank you very much for your answer. At present, I have tried using fully connected neural network, but the fitting accuracy has not reached my expectation. What to do

Maybe first looking into the problems that FCNN may have. Then if you cannot solve them, go for other architectures.

Dear Lo Show

Have a look:

1. https://kth.diva-portal.org/smash/get/diva2:1335815/FULLTEXT01.pdf

2. https://arxiv.org/pdf/1906.06543.pdf

Kind Regards

Qamar Ul Islam

thank you=Qamar Ul Islam&Mostafa Kiani Shahvandi