- outputCapture.PNG (79.55 KiB) Viewed 3284 times
Andrew,
I am using a setup very similar to your "TimeSeries" example.
Sometimes, after training I see a solution that is very biased - all results are to one side (usually positive) of zero. Yet I have scaled all my inputs, and my single output to -1,1 range.
If you can see the picture, the red dots represent the target output and the blue line is the solution.
What might cause such an obvious "bias"?
