arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1912.02427cs.LGcs.ITeess.SPmath.ITmath.OCstat.ML
收藏

过完备表示学习的优化景观分析

Analysis of the Optimization Landscapes for Overcomplete Representation Learning

New York University(纽约大学);UC Berkeley(加州大学伯克利分校);ByteDance Inc.(字节跳动公司);Chinese University of Hong Kong(香港中文大学);Rutgers University(罗格斯大学);Johns Hopkins University(约翰斯·霍普金斯大学);University of Denver(丹佛大学)

Qing Qu, Yuexiang Zhai, Xiao Li, Yuqian Zhang, Zhihui Zhu

摘要:

我们研究了学习过完备表示的非凸优化景观,包括学习(i)稀疏使用的过完备字典和(ii)卷积字典,这些无监督学习问题在高维数据分析中有许多应用。尽管简单的非凸算法在经验上取得了成功,但为什么这些方法工作得如此好的理论理由还远远不能令人满意。在这项工作中,我们证明了这些问题可以表示为$\ell^4$ -范数优化问题与球面约束,并研究了其非凸优化景观的几何性质。对于这两个问题,我们表明,非凸目标具有良性(全球)的几何结构,在这个意义上,每个局部极小值是接近的目标解决方案之一,每个鞍点表现出负曲率。这一发现使得发展保证全局优化方法,使用简单的初始化。对于这两个问题,我们表明非凸目标具有良性的几何结构-每个局部极小值接近的目标解决方案之一,每个鞍点表现出负曲率-无论是在整个空间或在一个足够大的区域。这一发现确保了具有简单初始化的局部搜索算法(例如黎曼梯度下降)近似地找到目标解。最后,数值实验证明了我们的理论发现。

英文摘要:

We study nonconvex optimization landscapes for learning overcomplete representations, including learning (i) sparsely used overcomplete dictionaries and (ii) convolutional dictionaries, where these unsupervised learning problems find many applications in high-dimensional data analysis. Despite the empirical success of simple nonconvex algorithms, theoretical justifications of why these methods work so well are far from satisfactory. In this work, we show these problems can be formulated as $\ell^4$-norm optimization problems with spherical constraint, and study the geometric properties of their nonconvex optimization landscapes. For both problems, we show the nonconvex objectives have benign (global) geometric structures, in the sense that every local minimizer is close to one of the target solutions and every saddle point exhibits negative curvature. This discovery enables the development of guaranteed global optimization methods using simple initializations. For both problems, we show the nonconvex objectives have benign geometric structures -- every local minimizer is close to one of the target solutions and every saddle point exhibits negative curvature -- either in the entire space or within a sufficiently large region. This discovery ensures local search algorithms (such as Riemannian gradient descent) with simple initializations approximately find the target solutions. Finally, numerical experiments justify our theoretical discoveries.

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。

↑