arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2209.07396stat.MLcs.LG
收藏

向消除分数匹配的盲目性迈进

Towards Healing the Blindness of Score Matching

University College London(伦敦大学学院)

Mingtian Zhang, Oscar Key, Peter Hayes, David Barber, Brooks Paige, François-Xavier Briol

AI总结:

本文提出了一种新的分歧家族,以缓解分数匹配在多模分布中的盲目性问题,并在密度估计中展示了改进的性能。

摘要:

基于分数的分歧已被广泛应用于机器学习和统计应用。尽管在经验上取得了成功,但在使用这些方法处理多模分布时却观察到了一个盲目性问题。在本文中,我们讨论了这个盲目性问题,并提出了一种新的分歧家族,可以缓解这个盲目性问题。我们将在密度估计的背景下展示所提出的分歧,并报告与传统方法相比有所改进的性能。

英文摘要:

Score-based divergences have been widely used in machine learning and statistics applications. Despite their empirical success, a blindness problem has been observed when using these for multi-modal distributions. In this work, we discuss the blindness problem and propose a new family of divergences that can mitigate the blindness problem. We illustrate our proposed divergence in the context of density estimation and report improved performance compared to traditional approaches.

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。

↑