arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1911.09189cs.LGcs.ITmath.ITstat.ML
收藏

无限宽神经网络的无限集成信息

Information in Infinite Ensembles of Infinitely-Wide Neural Networks

Edmond and Lilly Safra Center for Brain Sciences(埃德蒙与莉莉·萨夫拉脑科学中心);The Hebrew University of Jerusalem(耶路撒冷希伯来大学);Google Research(谷歌研究院)

Ravid Shwartz-Ziv, Alexander A. Alemi

摘要:

在这一初步工作中,我们研究了无限宽神经网络的无限集成的推广性质。令人惊讶的是,这个模型家族允许对许多信息论量进行易于处理的计算。我们报告的分析和实证调查,在寻找与泛化相关的信号。

英文摘要:

In this preliminary work, we study the generalization properties of infinite ensembles of infinitely-wide neural networks. Amazingly, this model family admits tractable calculations for many information-theoretic quantities. We report analytical and empirical investigations in the search for signals that correlate with generalization.

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。

↑