cs.LG（2023-10-28）

📊 共 11 篇论文 | 🔗 2 篇有代码

🎯 兴趣领域导航

#	题目	一句话要点	标签	🔗
1	Robust Offline Reinforcement learning with Heavy-Tailed Rewards	提出ROAM和ROOM以增强离线强化学习的鲁棒性	reinforcement learning offline RL offline reinforcement learning	✅
2	ReConTab: Regularized Contrastive Representation Learning for Tabular Data	提出ReConTab以解决表格数据特征工程的挑战	representation learning contrastive learning
3	Curriculum Learning for Graph Neural Networks: Which Edges Should We Learn First	提出基于课程学习的GNN边缘学习策略以提升表示能力	curriculum learning	✅
4	Temporally Disentangled Representation Learning under Unknown Nonstationarity	提出NCTRL框架以解决非平稳环境下的因果表示学习问题	representation learning
5	Using Early Readouts to Mediate Featural Bias in Distillation	提出早期读出机制以缓解蒸馏中的特征偏差问题	distillation
6	Weakly Coupled Deep Q-Networks	提出弱耦合深度Q网络以解决弱耦合马尔可夫决策过程问题	reinforcement learning deep reinforcement learning
7	Unsupervised Behavior Extraction via Random Intent Priors	提出UBER以从无奖励数据中提取多样化行为	reinforcement learning offline reinforcement learning

#	题目	一句话要点	标签
8	The Synergy of Speculative Decoding and Batching in Serving Large Language Models	提出自适应推测解码策略以提升大语言模型的GPU利用率	large language model
9	Large Language Models Are Better Adversaries: Exploring Generative Clean-Label Backdoor Attacks Against Text Classifiers	提出LLMBkd以解决文本分类器的清标签后门攻击问题	large language model
10	Efficient kernel surrogates for neural network-based regression	提出高效核代理以解决深度神经网络回归问题	foundation model
11	Optimal Transport for Kernel Gaussian Mixture Models	提出一种基于最优传输的核高斯混合模型距离度量	multimodal