Xiangyang Ji

Credit Assignment with Meta-Policy Gradient for Multi-Agent Reinforcement Learning

2021 · arXiv (Cornell University)

Reward decomposition is a critical problem in centralized training with decentralized execution~(CTDE) paradigm for multi-agent reinforcement learning. To take full advantage of global information, which exploits the states from all agents and the related environment …

Model-Free Reinforcement Learning: from Clipped Pseudo-Regret to Sample\n Complexity

2020 · arXiv (Cornell University)

In this paper we consider the problem of learning an $\\epsilon$-optimal\npolicy for a discounted Markov Decision Process (MDP). Given an MDP with $S$\nstates, $A$ actions, the discount factor $\\gamma \\in (0,1)$, and an\napproximation threshold $\\epsilon …

Local Manifold Augmentation for Multiview Semantic Consistency

2022 · arXiv (Cornell University)

Multiview self-supervised representation learning roots in exploring semantic consistency across data of complex intra-class variation. Such variation is not directly accessible and therefore simulated by data augmentations. However, commonly adopted augmentations are handcrafted and limited …

Backdoor Attacks Against Incremental Learners: An Empirical Evaluation Study

2023 · arXiv (Cornell University)

Large amounts of incremental learning algorithms have been proposed to alleviate the catastrophic forgetting issue arises while dealing with sequential data on a time series. However, the adversarial robustness of incremental learners has not been …