Yingbin Liang

Improving Sample Complexity Bounds for (Natural) Actor-Critic Algorithms

2020 · Neural Information Processing Systems

The actor-critic (AC) algorithm is a popular method to find an optimal policy in reinforcement learning. In the infinite horizon scenario, the finite-sample convergence rate for the AC and natural actor-critic (NAC) algorithms has been …

Provably Faster Algorithms for Bilevel Optimization

2021 · arXiv (Cornell University)

Bilevel optimization has been widely applied in many important machine learning applications such as hyperparameter optimization and meta-learning. Recently, several momentum-based algorithms have been proposed to solve bilevel optimization problems faster. However, those momentum-based algorithms …

Provable Generalization of Overparameterized Meta-learning Trained with SGD

2022 · arXiv (Cornell University)

Despite the superior empirical success of deep meta-learning, theoretical understanding of overparameterized meta-learning is still limited. This paper studies the generalization of a widely used meta-learning approach, Model-Agnostic Meta-Learning (MAML), which aims to find a …

Doubly Robust Instance-Reweighted Adversarial Training

2023 · arXiv (Cornell University)

Assigning importance weights to adversarial data has achieved great success in training adversarially robust networks under limited model capacity. However, existing instance-reweighted adversarial training (AT) methods heavily depend on heuristics and/or geometric interpretations to determine …