The CP tensor decomposition is a low-rank approximation of a tensor. We present a distributed-memory parallel algorithm and implementation of an alternating optimization method for computing a CP decomposition of dense tensors that can enforce nonnegativity of the computed low-rank factors. The principal task is to parallelize the Matricized-Tensor Times Khatri-Rao Product (MTTKRP) bottleneck subcomputation. The algorithm is computation efficient, using dimension trees to avoid redundant computation across MTTKRPs within the alternating method. Our approach is also communication efficient, using a data distribution and parallel algorithm across a multidimensional processor grid that can be tuned to minimize communication. We benchmark our software on synthetic as well as hyperspectral image and neuroscience dynamic functional connectivity data, demonstrating that our algorithm scales well to 100s of nodes (up to 4096 cores) and is faster and more general than the currently available parallel software.
more »
« less
PLANC: Parallel Low-rank Approximation with Nonnegativity Constraints
We consider the problem of low-rank approximation of massive dense nonnegative tensor data, for example, to discover latent patterns in video and imaging applications. As the size of data sets grows, single workstations are hitting bottlenecks in both computation time and available memory. We propose a distributed-memory parallel computing solution to handle massive data sets, loading the input data across the memories of multiple nodes, and performing efficient and scalable parallel algorithms to compute the low-rank approximation. We present a software package called Parallel Low-rank Approximation with Nonnegativity Constraints, which implements our solution and allows for extension in terms of data (dense or sparse, matrices or tensors of any order), algorithm (e.g., from multiplicative updating techniques to alternating direction method of multipliers), and architecture (we exploit GPUs to accelerate the computation in this work). We describe our parallel distributions and algorithms, which are careful to avoid unnecessary communication and computation, show how to extend the software to include new algorithms and/or constraints, and report efficiency and scalability results for both synthetic and real-world data sets.
more »
« less
- Award ID(s):
- 1642385
- NSF-PAR ID:
- 10276871
- Date Published:
- Journal Name:
- ACM Transactions on Mathematical Software
- Volume:
- 47
- Issue:
- 3
- ISSN:
- 0098-3500
- Page Range / eLocation ID:
- 1 to 37
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Over the last two decades, frameworks for distributed-memory parallel computation, such as MapReduce, Hadoop, Spark and Dryad, have gained significant popularity with the growing prevalence of large network datasets. The Massively Parallel Computation (MPC) model is the de-facto standard for studying graph algorithms in these frameworks theoretically. Subgraph counting is one such fundamental problem in analyzing massive graphs, with the main algorithmic challenges centering on designing methods which are both scalable and accurate. Given a graph G = (V, E) with n vertices, m edges and T triangles, our first result is an algorithm that outputs a (1+ε)-approximation to T, with asymptotically optimal round and total space complexity provided any S ≥ max{(√ m, n²/m)} space per machine and assuming T = Ω(√{m/n}). Our result gives a quadratic improvement on the bound on T over previous works. We also provide a simple extension of our result to counting any subgraph of k size for constant k ≥ 1. Our second result is an O_δ(log log n)-round algorithm for exactly counting the number of triangles, whose total space usage is parametrized by the arboricity α of the input graph. We extend this result to exactly counting k-cliques for any constant k. Finally, we prove that a recent result of Bera, Pashanasangi and Seshadhri (ITCS 2020) for exactly counting all subgraphs of size at most 5 can be implemented in the MPC model in Õ_δ(√{log n}) rounds, O(n^δ) space per machine and O(mα³) total space. In addition to our theoretical results, we simulate our triangle counting algorithms in real-world graphs obtained from the Stanford Network Analysis Project (SNAP) database. Our results show that both our approximate and exact counting algorithms exhibit improvements in terms of round complexity and approximation ratio, respectively, compared to two previous widely used algorithms for these problems.more » « less
-
Persistent homology (PH) matrix reduction is an important tool for data analytics in many application areas. Due to its highly irregular execution patterns in computation, it is challenging to gain high efficiency in parallel processing for increasingly large data sets. In this paper, we introduce HYPHA, a HYbrid Persistent Homology matrix reduction Accelerator, to make parallel processing highly efficient on both GPU and multicore. The essential foundation of our algorithm design and implementation is the separation of SIMT and MIMD parallelisms in PH matrix reduction computation. With such a separation, we are able to perform massive parallel scanning operations on GPU in a super-fast manner, which also collects rich information from an input boundary matrix for further parallel reduction operations on multicore with high efficiency. The HYPHA framework may provide a general purpose guidance to high performance computing on multiple hardware accelerators. To our best knowledge, HYPHA achieves the highest performance in PH matrix reduction execution. Our experiments show speedups of up to 116x against the standard PH algorithm. Compared to the state-of-the-art parallel PH software packages, such as PHAT and DIPHA, HYPHA outperforms their fastest PH matrix reduction algorithms by factor up to 2.3x.more » « less
-
Graph processing recently received intensive interests in light of a wide range of needs to understand relationships. It is well-known for the poor locality and high memory bandwidth requirement. In conventional architectures, they incur a significant amount of data movements and energy consumption which motivates several hardware graph processing accelerators. The current graph processing accelerators rely on memory access optimizations or placing computation logics close to memory. Distinct from all existing approaches, we leverage an emerging memory technology to accelerate graph processing with analog computation. This paper presents GRAPHR, the first ReRAM-based graph processing accelerator. GRAPHR follows the principle of near-data processing and explores the opportunity of performing massive parallel analog operations with low hardware and energy cost. The analog computation is suitable for graph processing because: 1) The algorithms are iterative and could inherently tolerate the imprecision; 2) Both probability calculation (e.g., PageRank and Collaborative Filtering) and typical graph algorithms involving integers (e.g., BFS/SSSP) are resilient to errors. The key insight of GRAPHR is that if a vertex program of a graph algorithm can be expressed in sparse matrix vector multiplication (SpMV), it can be efficiently performed by ReRAM crossbar. We show that this assumption is generally true for a large set of graph algorithms. GRAPHR is a novel accelerator architecture consisting of two components: memory ReRAM and graph engine (GE). The core graph computations are performed in sparse matrix format in GEs (ReRAM crossbars). The vector/matrix-based graph computation is not new, but ReRAM offers the unique opportunity to realize the massive parallelism with unprecedented energy efficiency and low hardware cost. With small subgraphs processed by GEs, the gain of performing parallel operations overshadows the wastes due to sparsity. The experiment results show that GRAPHR achieves a 16.01X (up to 132.67X) speedup and a 33.82X energy saving on geometric mean compared to a CPU baseline system. Compared to GPU, GRAPHR achieves 1.69X to 2.19X speedup and consumes 4.77X to 8.91X less energy. GRAPHR gains a speedup of 1.16X to 4.12X, and is 3.67X to 10.96X more energy efficiency compared to PIM-based architecture.more » « less
-
Although state-of-the-art (SOTA) CNNs achieve outstanding performance on various tasks, their high computation demand and massive number of parameters make it difficult to deploy these SOTA CNNs onto resource-constrained devices. Previous works on CNN acceleration utilize low-rank approximation of the original convolution layers to reduce computation cost. However, these methods are very difficult to conduct upon sparse models, which limits execution speedup since redundancies within the CNN model are not fully exploited. We argue that kernel granularity decomposition can be conducted with low-rank assumption while exploiting the redundancy within the remaining compact coefficients. Based on this observation, we propose PENNI, a CNN model compression framework that is able to achieve model compactness and hardware efficiency simultaneously by (1) implementing kernel sharing in convolution layers via a small number of basis kernels and (2) alternately adjusting bases and coefficients with sparse constraints. Experiments show that we can prune 97% parameters and 92% FLOPs on ResNet18 CIFAR10 with no accuracy loss, and achieve 44% reduction in run-time memory consumption and a 53% reduction in inference latency.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1145/3432185
