作者提出 t-SVDM-I 与 t-SVDM-II 的首个分布式内存实现,面向 N 维张量,并用异步非阻塞通信优化数据重分布与关键计算内核。
cs.DC
查看原文摘要
The star-M tensor algebra generalizes the optimal compression guarantees of the matrix singular value decomposition (SVD) to higher-order tensors. The t-SVDM-I decomposition is computed using a set of invertible transformations and slice-wise truncated SVDs with a prescribed rank, and its adaptive, multirank variant t-SVDM-II can determine slice-wise ranks to satisfy a prescribed error tolerance. Existing implementations of these algorithms are limited to shared-memory environments, restricting their applicability to modern large-scale datasets. In this work, we present the first distributed-memory implementations of t-SVDM-I and t-SVDM-II, which are generalized to N-way tensors. To address computational and communication bottlenecks, we design asynchronous, non-blocking communication strategies for efficient data redistribution to obtain highly efficient key computational kernels. Strong-scaling experiments demonstrate the efficiency and scalability of our approach (at least 70% parallel efficiency on up to 128 nodes, or 6144 cores), and we apply the star-M compression techniques to a set of application datasets of size up to 768 GB.
DuBose Tuller, Grey Ballard, Hussam Al Daas, Nicholas I. M. Gould
原文摘要
We present an algorithm for computing the CP tensor decomposition using Newton's method. The problem of fitting a CP model to a data tensor with least squares loss can be solved by various gradient-based optimization techniques, including Gauss-Newton methods that exploit the tensor structure of the problem and iteratively solve the linear system involving the approximate Hessian matrix to determine the step direction at each iteration. Gauss-Newton methods exhibit superlinear convergence when the residual is small but can otherwise struggle to converge when the approximation to the Hessian is not accurate. We show that the Gauss-Newton approach can be extended to Newton's method, using the true Hessian to obtain fast convergence even in the case of large residuals. In particular, we derive efficient means of applying the true Hessian to vectors for use within an iterative solver, we employ globalization and regularization techniques to make the Newton model robust to ill-conditioned and/or indefinite Hessians, and we propose hybrid use of Gauss-Newton and Newton steps to balance convergence behavior both far and near local minima. Our numerical experiments use synthetic examples to show in which scenarios Gauss-Newton and other methods struggle to converge, and we demonstrate the superiority of the hybrid approach on two application data sets.