LSTM different results


When I apply my LSTM to obtain two outputs (one is symmetric to the other), my LSTM does not perform the same in both outputs, one is clearly better than the other.


Has anyone had the same problem?

on what data was it trained?

My dataset consists on:
92 different dimensions of transistors
For each dimension, I have 9 datasets with 2 input waves and 2 output waves.

The objective is to mimic the behaviour of the circuit for all the dimensions, in other words it need to pe capable to, given a dimension and the input waves, produce the correct two output waves