Rather than approximating the model, counterfactuals are promising since they always convey an accurate picture about the ML model. We outline our method for generating multiple diverse CFs in our #FAT2020 paper: https://arxiv.org/abs/1905.07697 with @ChenhaoTan & Ramaravind Mothilal
-
-
Prikaži ovu nit
-
Many challenges still remain. Perhaps the biggest is how to satisfy causal constraints when generating counterfactuals, so that they are actionable. E.g., it is impossible to change one's education without aging. Some preliminary work https://arxiv.org/abs/1912.03277 3/3
Prikaži ovu nit
Kraj razgovora
Novi razgovor -
-
-
Super cool, congrats! Would be fun to look into integrating use of this library into our visual What-If Tool. Could chat at FAT* tomorrow if you or co-authors are around.
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
That is a simple idea, very cool
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.