Concatenate the input vector of the NN a constant vector

1

Hi.

I am trying to invert a physical model in which constant data is used as a vector. I would like to input both the model output and this constant vector data to the NN.
How can I do this?

Thanks and regards …

2

Have you tried just using torch.cat like this:

input = torch.cat([out, input])
3

Picture1
Picture11719×414 29.1 KB

I don’t have much experience in pytorch or neural networks, but from what I understand, if I concatenate the constant vector, I have to do it to each of the inputs of the training set, so I multiply by 4 the memory, and since I am using a very large set of training, I think that is not an option.
I have drawn a diagram of the inputs and outputs of the model, I use it to generate the training set, as you can see the model uses only the three vectors of size 80.
I wonder if there is any way to introduce this information in the NN to invert the model and without quadrupling the information of the training set.