Temporal Convoutional Networks for Regression

Hi,

I am trying to construct a TCN for regression purposes with only one output. I found this on matlab support which is an excellent example.

What changes would I need to make to these so that I can convert it to a regression scheme that only cares about the last output?

Additionally, would it be better to do 1D CNN for this?

My input is a long time series and the output is a scalar that defines a certain property of the time series.

Thank you in advance.

Kind Regards,

[틈 마]

I want to do the same work like you to learn TCN, How about your coding?

[Yeray Pabon]

I'm definitely trying to do this but I keep getting the error that the response must be the same as the sequence length, which doesn't make sense when doing regression with a single output.

Is there anyone that can provide an example or a fix for this?

@MathWorks Support Team

[KSSV]

@Yeray Pabon Attach your code and data.

[Chris Lehnig]

Did you try swabbing the classificationLayer with a regressionLayer?