Timezone: »

Understanding Certified Training with Interval Bound Propagation
Yuhao Mao · Mark Müller · Marc Fischer · Martin Vechev

Fri Jul 28 02:30 PM -- 03:00 PM (PDT) @

As robustness verification methods are becoming more precise, training certifiably robust neural networks is becoming ever more relevant. To this end, certified training methods compute and then optimize an upper bound on the worst-case loss over a robustness specification. Curiously, training methods based on the imprecise interval bound propagation (IBP) consistently outperform those leveraging more precise bounding methods. Still, we lack an understanding of the mechanisms making IBP so successful.In this work, we thoroughly investigate these mechanisms theoretically and empirically by leveraging a novel metric measuring the tightness of IBP bounds.

Author Information

Yuhao Mao (ETH Zurich)
Mark Müller (ETH Zurich)
Marc Fischer (ETH Zurich)
Martin Vechev (ETH Zurich)

More from the Same Authors