Vector Researchers Win Awards at ICLR 2022 Conference
Researchers from the Vector Institute for Artificial Intelligence achieved significant recognition at the 2022 International Conference on Learning Representations (ICLR). Two papers co-authored by Vector faculty members received Outstanding Paper Awards. Nicolas Papernot and Thomas Steinke were honored for their work on Hyperparameter Tuning with Renyi Differential Privacy, which addresses privacy leakage during algorithm training and hyperparameter tuning. Additionally, Vardan Papyan, along with co-authors X.Y. Han and David L. Donoho, received an award for their study on Neural Collapse under MSE Loss, revealing geometric structures in deep network training that offer insights into adversarial robustness and generalization. In total, 21 papers involving Vector Faculty Members and Affiliates were accepted to the virtually held conference. Among these, two were selected for oral presentations and five for spotlight talks. Other notable contributions included research on architecture-independent model distances using LIME and accelerated policy learning with differentiable simulation. This achievement underscores the institute's prominent role in advancing artificial intelligence research and its impact on the global machine learning community.
Wire timeline
Vector Researchers Win Awards at ICLR 2022 Conference
Researchers from the Vector Institute for Artificial Intelligence achieved significant recognition at the 2022 International Conference on Learning Representations (ICLR). Two papers co-authored by Vector faculty members received Outstanding Paper Awards. Nicolas Papernot and Thomas Steinke were honored for their work on Hyperparameter Tuning with Renyi Differential Privacy, which addresses privacy leakage during algorithm training and hyperparameter tuning. Additionally, Vardan Papyan, along with co-authors X.Y. Han and David L. Donoho, received an award for their study on Neural Collapse under MSE Loss, revealing geometric structures in deep network training that offer insights into adversarial robustness and generalization. In total, 21 papers involving Vector Faculty Members and Affiliates were accepted to the virtually held conference. Among these, two were selected for oral presentations and five for spotlight talks. Other notable contributions included research on architecture-independent model distances using LIME and accelerated policy learning with differentiable simulation. This achievement underscores the institute's prominent role in advancing artificial intelligence research and its impact on the global machine learning community.
Vector Institute for Artificial Intelligence