I’m trying to construct minimal neural networks for a set of functions. the functions are complete – I’m trying to replicate them completely in NN, with minimal resources (non-zero weights).
(Obviously the term “minimal” is very loose, there will be heavy constraints on the contents/structure of the network, but im fine with that)

I built a 4-layer feed forward network. However the prediction for the test sample are all the same.They are a fixed number. I have standardized the data into [-1,1]. I am not sure if there is something wrong with my code.

I am trying to set up a physics-informed neural network in Python with deepXDE using a pytorch backend. My problem is quite simple and consists of only 2 points in the domain with 3 inputs (time, space, area) and 1 output (pressure).

My neural network always predicts the same value, the problem arises when I try to normalize the target value, if I don’t it works fine. Does anyone have an idea of why?

Relatively recently, as part of my research work, I had to numerically solve a system of differential equations, and instead of using ready-made programs written in Fortran, I decided to write my own in Python. As one of the intermediate steps, I encountered a system of differential equations, which you can see in the attachment. I found a module that solves differential equations using machine learning, and it seemed to me that it would be suitable for solving this particular problem. But a question arose. How do I specify all the boundary conditions? The ones that are set straightforwardly as specific numbers are easy to specify, but what should I do with those where the relationship between the derivative of the function and the function itself is specified?

I’m facing an issue with the train function in my neural network implementation. Below is the relevant code snippet:

I’m coding my very first neural network model in python. The data I use for training has 14 features, 4 different classes and a total of 1953 instances. I use mini-batches of 4 instances for training. However, the error only keeps increasing with each training epoch, even when I adjust the learning rate. I will add the methods that I believe might cause the issue below. This is the graph I get for 60 training epochs with a learning rate of 0.001 (changing the learning rate doesn’t change the overall shape of the graph) :