The current NARX network seems only supporting 1 output only

[tzhu-ece]

In narx_nn.py, line 374

It seems this line is limiting output to single value, can you make an update to support multiple outputs?

[wilsonrljr]

Hi @tzhu-ece , thanks for commenting. Currently we only support MISO systems. We have a core function to handle the lagged features and we have to make some modifications and tests for MIMO problems. I'll take a look to check how I could release the modification for MIMO systems and I update you here.

[jifei-deng]

Great project! Did you complete the modification for MIMO?

[jankap]

I'd be very interested in that feature, too. Can we assist somehow? Thank you very much for your awesome work

[wilsonrljr]

Hi! I didn't complete this yet. I've just published the v0.1.9 update and I'll start working in the narx neural net module. First I will implement the n-steps ahead forecast (currently its support only infinity steps ahead), update the code following the update 0.1.7 and I'll take a look at this request.

Soon I'll be back with some questions regarding specific details about the MIMO feature like if one output depends or not on other output.

Any help will be very appreciated. If you guys have any dataset so I can test on it and use as benchmark, just let me know.

[jankap]

questions regarding specific details about the MIMO feature like if one output depends or not on other output

that should be configurable like in ML System Identification Toolbox imo, where you can choose custom regressors: https://www.mathworks.com/help/ident/ref/customregressor.html
The issue: These custom regressors are rather slow in ML :D