近日，ICML 2022 大会在美国马里兰州巴尔的摩市以线上线下结合的方式举办。这也是新冠疫情以来大会首次恢复线下形式举办。
目前，大会已经公布了全部奖项，包括 15 篇杰出论文奖和 1 项时间检验奖。其中，复旦大学、上海交通大学、厦门大学、莱斯大学胡侠团队等多个华人团队的研究获得杰出论文奖。ICML 2012 关于「投毒攻击」的论文《Poisoning Attacks against Support Vector Machines》获得了本次大会的时间检验奖。

杰出论文奖：

1、Monarch: Expressive Structured Matrices for Efficient and Accurate Training
2、Solving Stackelberg Prediction Game with Least Squares Loss via Spherically Constrained Least Squares Reformulation
3、G-Mixup: Graph Data Augmentation for Graph Classification
4、Stable Conformal Prediction Sets
5、Learning inverse folding from millions of predicted structures
6、Causal Conceptions of Fairness and their Consequences
7、Do Differentiable Simulators Give Better Policy Gradients?
8、Learning Mixtures of Linear Dynamical Systems
9、Active fairness auditing
10、Understanding Dataset Difficulty with V-Usable Information
11、Adversarially Trained Actor Critic for Offline Reinforcement Learning
12、Privacy for Free: How does Dataset Condensation Help Privacy?
13、Bayesian Model Selection, the Marginal Likelihood, and Generalization
14、The Importance of Non-Markovianity in Maximum State Entropy Exploration
15、Minimum Cost Intervention Design for Causal Effect Identification

 

时间检验奖：

论文题目：Poisoning Attacks against Support Vector Machines