arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1810.03218cs.LGmath.OCstat.ML
收藏

基于线性规划的深度神经网络训练

Principled Deep Neural Network Training through Linear Programming

Columbia University(哥伦比亚大学);Universidad de O’Higgins(奥希金斯大学);Zuse Institute Berlin(柏林祖斯研究所)

Daniel Bienstock, Gonzalo Muñoz, Sebastian Pokutta

摘要:

深度学习最近受到了广泛关注,因为训练算法取得了令人印象深刻的经验表现。因此,需要更好地从理论上理解这些问题,近年来变得更加明显。在这项工作中,使用一个统一的框架,我们表明,存在一个多面体,同时编码所有可能的深度神经网络训练问题,可以产生一个给定的架构,激活函数,损失函数,和样本大小。值得注意的是,多面体表示的大小仅线性地依赖于样本大小,并且不太可能更好地依赖于其他几个网络参数(假设$P\neq NP$)。此外,我们使用我们的多面体表示,以获得新的和更好的计算复杂性的结果,为训练问题的知名神经网络架构。我们的研究结果提供了一个新的视角,通过多面体理论的镜头训练问题,并揭示了一个强大的结构所产生的这些问题。

英文摘要:

Deep learning has received much attention lately due to the impressive empirical performance achieved by training algorithms. Consequently, a need for a better theoretical understanding of these problems has become more evident in recent years. In this work, using a unified framework, we show that there exists a polyhedron which encodes simultaneously all possible deep neural network training problems that can arise from a given architecture, activation functions, loss function, and sample-size. Notably, the size of the polyhedral representation depends only linearly on the sample-size, and a better dependency on several other network parameters is unlikely (assuming $P\neq NP$). Additionally, we use our polyhedral representation to obtain new and better computational complexity results for training problems of well-known neural network architectures. Our results provide a new perspective on training problems through the lens of polyhedral theory and reveal a strong structure arising from these problems.

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。

↑