Salon with UC Berkeley's ML@B
Teams from UC Berkeley's ML@B student organization present their recent projects on gesture detection, capsule networks, and adversarial attacks.
This salon features teams from the ML@B
student organization at the University of California, Berkeley, presenting their work on ML research and engineering over the past semester.
Video Gesture Overlay, by Ashwat Chidambaram, Andre He, Sarvasv Kulpati, and Aryia Dattamajumdar. This team put together a machine learning-based video overlay that uses hand gestures to trigger emoji reactions (and more!) on Zoom.
Semantic Convolutions, by Domas Buracas, Nitish Dashora, Kenny Workman, and Max Smolin. The goal of the project is to investigate how to make CNNs more interpretable. They'll present work on Capsule Networks -- including implementing them in PyTorch and finding speed-ups.
Principled Attacks on Adversarial Example Detectors, by Oliver Bryniarski, Nabeel Hingun, Pedro Pachuca, and Vincent Wang (advised by Nicholas Carlini of Google Brain). This group will discuss some of the classical techniques and results in adversarial attack and defense, then talk about attacking error correction-based defense schemes.