Performance decreases when open loop NARX simulated closed loop

Hi everyone,

My network performance decreases very much when I train it open loop but then simulate closed loop.

Instead performance is far better when I train it close loop as well. It is not recommended to be done in that way just because it takes long time or there is some other reason that is not correct?

Thank you in advance,

Esra

[FRANCISCO]

Hello esra, As he had shown me the network performance in closeloop training was much lower than if the train in and then do closeloop openloop. I guess it also depends on the correlation showing your variables.

[esra]

But in the handbook and documents it says that it is better to train open loop and then close the loop. I am confused..