Timezone: »

 
Poster
Contextual Reliability: When Different Features Matter in Different Contexts
Gaurav Ghosal · Amrith Setlur · Daniel S Brown · Anca Dragan · Aditi Raghunathan

Thu Jul 27 01:30 PM -- 03:00 PM (PDT) @ Exhibit Hall 1 #115
in Poster Session 5 »

Deep neural networks often fail catastrophically by relying on spurious correlations. Most prior work assumes a clear dichotomy into spurious and reliable features; however, this is often unrealistic. For example, most of the time we do not want an autonomous car to simply copy the speed of surrounding cars---we don't want our car to run a red light if a neighboring car does so. However, we cannot simply enforce invariance to next-lane speed, since it could provide valuable information about an unobservable pedestrian at a crosswalk. Thus, universally ignoring features that are sometimes (but not always) reliable can lead to non-robust performance. We formalize a new setting called contextual reliability which accounts for the fact that the "right" features to use may vary depending on the context. We propose and analyze a two-stage framework called Explicit Non-spurious feature Prediction (ENP) which first identifies the relevant features to use for a given context, then trains a model to rely exclusively on these features. Our work theoretically and empirically demonstrates the advantages of ENP over existing methods and provides new benchmarks for contextual reliability.

[ PDF]  [ Poster] 

Author Information

Gaurav Ghosal (University of California Berkeley)
Amrith Setlur (Carnegie Mellon University)
Daniel S Brown (University of Utah)
Anca Dragan (University of California, Berkeley)
Aditi Raghunathan (Carnegie Mellon University)

More from the Same Authors