What are the vulnerabilities of machine learning models, what are the latest types of data poisoning attacks, and what are the legal issues pertaining to creative artificial intelligence: This is just a very small sample of the questions that researchers from around the world will be discussing at the International Conference on Machine Learning, which is being held in sunny Hawaii this year.
In addition to poster sessions and paper presentations, many workshops will be attracting members of the international research community. One of these workshops is the "2nd Workshop on Formal Verification of Machine Learning", co-organized by members of ELSA – European Lighthouse on Secure and Safe AI. An EU-funded virtual network of excellence, ELSA is coordinated by CISPA-Faculty Professor Dr. Mario Fritz. ELSA is presenting this year’s "Outstanding Paper Awards", an honor bestowed on the most impressive of all submitted papers.
The ICML accepted the following eight papers involving CISPA researchers:
- “Why Random Pruning Is All We Need to Start Sparse” by Advait Gadhikar, Sohom Mukherjee, Rebekka Burkholz
- “On the Relationship Between Explanation and Prediction: A Causal View, Amir-Hossein Karimi” by Krikamol Muandet, Simon Kornblith, Bernhard Schölkopf, Been Kim
- “Nonlinear Causal Discovery with Latent Confounders” by David Kaltenpoth, Jilles Vreeken
- “Generated Graph Detection” by Yihan Ma, Zhikun Zhang, Ning Yu, Xinlei He, Michael Backes, Yun Shen, Yang Zhang
- “Data Poisoning Attacks Against Multimodal Encoders” by Ziqing Yang, Xinlei He, Zheng Li, Michael Backes, Mathias Humbert, Pascal Berrang, Yang Zhang
- “Special Properties of Gradient Descent with Large Learning Rates” by Amirkeivan Mohtashami, Martin Jaggi, Sebastian Stich
- “Revisiting Gradient Clipping: Stochastic bias and tight convergence guarantees” by Anastasiia Koloskova, Hadrien Hendrikx, Sebastian Stich
- “Conformal Prediction Sets for Graph Neural Networks” by Soroush H. Zargarbashi, Simone Antonelli, Aleksandar Bojchevski
