arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2403.07282cs.LG
收藏

通过灵活的非参数后验采样增强迁移学习

Enhancing Transfer Learning with Flexible Nonparametric Posterior Sampling

The University of Hong Kong(香港大学);KAIST AI(韩国科学技术院人工智能);AITRICS(AITRICS公司)

Hyungi Lee, Giung Nam, Edwin Fong, Juho Lee

AI总结:

针对迁移学习中上下游数据分布偏移导致现有先验策略效果受限的问题,提出采用非参数先验的NPTL后验采样方法,经实验验证其BMA性能优于其他基准。

摘要:

迁移学习近期在涉及深度神经网络的各类任务中展现出优异性能。在这类迁移学习场景下,下游数据的先验分布在贝叶斯模型平均(Bayesian model averaging, BMA)中至关重要。已有研究提出以预训练解为中心设定神经网络参数的先验,但这类策略在应对上游与下游数据间的分布偏移时存在局限。本文提出非参数迁移学习(nonparametric transfer learning, NPTL),这是一种灵活的后验采样方法,用于在非参数学习框架下解决分布偏移问题。非参数学习(nonparametric learning, NPL)是近期提出的方法,采用非参数先验进行后验采样,能有效处理模型误设场景,适用于可能存在上下游任务分布偏移的迁移学习场景。通过大量实证验证,我们证明该方法的BMA性能优于其他基准方法。

英文摘要:

Transfer learning has recently shown significant performance across various tasks involving deep neural networks. In these transfer learning scenarios, the prior distribution for downstream data becomes crucial in Bayesian model averaging (BMA). While previous works proposed the prior over the neural network parameters centered around the pre-trained solution, such strategies have limitations when dealing with distribution shifts between upstream and downstream data. This paper introduces nonparametric transfer learning (NPTL), a flexible posterior sampling method to address the distribution shift issue within the context of nonparametric learning. The nonparametric learning (NPL) method is a recent approach that employs a nonparametric prior for posterior sampling, efficiently accounting for model misspecification scenarios, which is suitable for transfer learning scenarios that may involve the distribution shift between upstream and downstream tasks. Through extensive empirical validations, we demonstrate that our approach surpasses other baselines in BMA performance.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑