Open alitirmizi23 opened 1 year ago
@alitirmizi23 hello! have you found a way to do this?
Hi @dyhan316 and @alitirmizi23,
Unfortunatly, this is not possible with the current methods in TSInterpret. There is literature on how to do this for Regression. However for now they are not implemented here. (So far I did not have the time). I still left the issue open, as it is a relevant addition and I plan on adding some algorithms when time allows it.
If you are interested, I can point you to some relevant Literature.
Also feel free to contribute :)
Hi @JHoelli,
I would also be interested in such a feature. I would like to build an XAI TCN model for univariate TS regression, so I would love to have read to Literature you mentioned.
Hi @AlexisTabin,
i started a summary repository for the topic "(Post-Hoc) Explainability for time series" here with a subsection for regression / forecating. Feel free to look through the sources and linked github repos or add things if you feel something is missing. I will also extend the list at some point.
My best guess would be trying Dynamask (github, paper).
I hope that helps.
Motivation. and
Describe the solution you'd like similar functionality as is available for time series classification, should be available for time series regression