ICML 2020 那些不容忽视的高产华人学者
Percy Liang 本次共有 8 篇文章入选,位居华人贡献榜首位。
Percy Liang 是斯坦福大学计算机科学副教授(麻省理工学士,2004 年;加州大学伯克利分校博士,2011 年)。他的两个研究目标是(i)使机器学习更加健壮、公平和可解释;以及(ii)使计算机更易于通过自然语言进行交流。他的奖项包括总统科学家和工程师早期职业奖(2019 年)、IJCAI计算机和思想奖(2016 年)、NSF职业奖(2016 年)、斯隆研究奖学金(2015 年)和微软研究学院奖学金(2014 年),此外,他还是 ICML2017 年最佳论文奖得主。
入选论文:
1 Understanding and Mitigating the Tradeoff Between Robustness and Accuracy
2 Understanding Self-Training for Gradual Domain Adaptation
3 Feature Noise Induces Loss Discrepancy Across Groups
4 Robustness to Spurious Correlations via Human Annotations
5 Adversarial Mutual Information for Text Generation
6 Concept Bottleneck Models
7 Overparameterization hurts worst-group accuracy with spurious correlations
8 DrRepair: A Self-Supervised, Graph-Attentional Approach to Repairing Programs from Diagnostic Feedback
Zhaoran Wang 是美国西北大学工程与管理系的助理教授,他在普林斯顿大学获得了博士学位。研究兴趣为统计优化与机器学习的交互。特别有兴趣了解统计准确性和计算工作量之间的权衡,以及为统计学习中的非凸问题建立可证明的保证。
入选论文:
1 Semiparametric Nonlinear Bipartite Graph Representation Learning with Provable Guarantees
2 Provably Efficient Exploration in Policy Optimization
3 Deep Reinforcement Learning with Smooth Policy
4 Breaking the Curse of Many Agents: Provable Mean Embedding QQQ-Iteration for Mean-Field Reinforcement Learning
5 Generative Adversarial Imitation Learning with Neural Network Parameterization: Global Optimality and Convergence Rate
6 Computational and Statistical Tradeoffs in Inferring Combinatorial Structures of Ising Model
7 On the Global Optimality of Model-Agnostic Meta-Learning
Chi Jin(金驰)以 6 篇论文入选位居华人贡献榜第3位。
Chi Jin 现任普林斯顿大学电气工程助理教授。他在加州大学伯克利分校获得计算机科学博士学位,由 Michael I. Jordan 担任导师。在此之前,他在北京大学获得了物理学学士学位。研究兴趣在于机器学习、统计和优化。
入选论文:
1 On Gradient Descent Ascent for Nonconvex-Concave Minimax Problems
2 What is Local Optimality in Nonconvex-Nonconcave Minimax Optimization?
3 Provably Efficient Exploration in Policy Optimization
4 Learning Adversarial Markov Decision Processes with Bandit Feedback and Unknown Transition
5 Reward-Free Exploration for Reinforcement Learning
6 Provable Self-Play Algorithms for Competitive Reinforcement Learning
1 Semiparametric Nonlinear Bipartite Graph Representation Learning with Provable Guarantees
2 Provably Efficient Exploration in Policy Optimization
3 Breaking the Curse of Many Agents: Provable Mean Embedding QQQ-Iteration for Mean-Field Reinforcement Learning
4 Generative Adversarial Imitation Learning with Neural Network Parameterization: Global Optimality and Convergence Rate
5 Robust One-Bit Recovery via ReLU Generative Networks: Near Optimal Statistical Rate and Global Landscape Analysis
6 On the Global Optimality of Model-Agnostic Meta-Learning
Heng Huang(黄恒)以 5 篇文章入选位居华人贡献榜第 4。
Heng Huang(黄恒)在达特茅斯学院获得计算机科学博士学位,在上海交通大学获得硕士和学士学位。他是匹兹堡大学(universityofpittsburgh)计算机工程 johna.Jurenko 特聘教授。研究兴趣为机器学习,数据挖掘,大数据计算,NLP 生物信息学,神经信息学,精密医学,健康信息学计算机视觉、医学图像分析。
入选论文:
1 Sparse Shrunk Additive Models
2 Fast OSCAR and OWL with Safe Screening Rules
3 Momentum-Based Policy Gradient Methods
4 Can Stochastic Zeroth-Order Frank-Wolfe Method Converge Faster for Non-Convex Problems
5 Adversarial Nonnegative Matrix Factorization
ICML 涌现的学术新星
以下同学有 2 篇论文入选: