arXiv subjects

John D. Owens

Publications and source records attributed to John D. Owens.

At least 37 records · Page 2Linked to original sources

arXiv2021

Energy-based Out-of-distribution Detection

Determining whether inputs are out-of-distribution (OOD) is an essential building block for safely deploying machine learning models in the open world. However, previous methods relying on the softmax confidence score suffer from overconfident posterior distributions for OOD data. We propose a unified framework for OOD detection that uses an energy score. We show that energy scores better distinguish in- and out-of-distribution samples than the traditional approach using the softmax scores. Unlike softmax confidence scores, energy scores are theoretically aligned with the probability density of the inputs and are less susceptible to the overconfidence issue. Within this framework, energy can be flexibly used as a scoring function for any pre-trained neural classifier as well as a trainable cost function to shape the energy surface explicitly for OOD detection. On a CIFAR-10 pre-trained WideResNet, using the energy score reduces the average FPR (at TPR 95%) by 18.03% compared to the softmax confidence score. With energy-based training, our method outperforms the state-of-the-art on common benchmarks.

cs.LG↗

arXiv2020

Fast Gunrock Subgraph Matching (GSM) on GPUs

In this paper, we propose a GPU-efficient subgraph isomorphism algorithm using the Gunrock graph analytic framework, GSM (Gunrock Subgraph Matching), to compute graph matching on GPUs. In contrast to previous approaches on the CPU which are based on depth-first traversal, GSM is BFS-based: possible matches are explored simultaneously in a breadth-first strategy. The advantage of using BFS-based traversal is that we can leverage the massively parallel processing capabilities of the GPU. The disadvantage is the generation of more intermediate results. We propose several optimization techniques to cope with the problem. Our implementation follows a filtering-and-verification strategy. While most previous work on GPUs requires one-/two-step joining, we use one-step verification to decide the candidates in current frontier of nodes. Our implementation has a speedup up to 4x over previous GPU state-of-the-art implementation.

cs.DC↗

arXiv2019

Object Localization with a Weakly Supervised CapsNet

Inspired by CapsNet's routing-by-agreement mechanism with its ability to learn object properties, we explore if those properties in turn can determine new properties of the objects, such as the locations. We then propose a CapsNet architecture with object coordinate atoms and a modified routing-by-agreement algorithm with unevenly distributed initial routing probabilities. The model is based on CapsNet but uses a routing algorithm to find the objects' approximate positions in the image coordinate system. We also discussed how to derive the property of translation through coordinate atoms and we show the importance of sparse representation. We train our model on the single moving MNIST dataset with class labels. Our model can learn and derive the coordinates of the digits better than its convolution counterpart that lacks a routing-by-agreement algorithm, and can also perform well when testing on the multi-digit moving MNIST and KTH datasets. The results show our method reaches the state-of-art performance on object localization without any extra localization techniques and modules as in prior work.

cs.CV↗

arXiv2019

Unsupervised Object Segmentation with Explicit Localization Module

In this paper, we propose a novel architecture that iteratively discovers and segments out the objects of a scene based on the image reconstruction quality. Different from other approaches, our model uses an explicit localization module that localizes objects of the scene based on the pixel-level reconstruction qualities at each iteration, where simpler objects tend to be reconstructed better at earlier iterations and thus are segmented out first. We show that our localization module improves the quality of the segmentation, especially on a challenging background.

cs.CV↗

arXiv2019

RDMA vs. RPC for Implementing Distributed Data Structures

Distributed data structures are key to implementing scalable applications for scientific simulations and data analysis. In this paper we look at two implementation styles for distributed data structures: remote direct memory access (RDMA) and remote procedure call (RPC). We focus on operations that require individual accesses to remote portions of a distributed data structure, e.g., accessing a hash table bucket or distributed queue, rather than global operations in which all processors collectively exchange information. We look at the trade-offs between the two styles through microbenchmarks and a performance model that approximates the cost of each. The RDMA operations have direct hardware support in the network and therefore lower latency and overhead, while the RPC operations are more expressive but higher cost and can suffer from lack of attentiveness from the remote side. We also run experiments to compare the real-world performance of RDMA- and RPC-based data structure operations with the predicted performance to evaluate the accuracy of our model, and show that while the model does not always precisely predict running time, it allows us to choose the best implementation in the examples shown. We believe this analysis will assist developers in designing data structures that will perform well on current network architectures, as well as network architects in providing better support for this class of distributed data structures.

cs.DC↗

arXiv2019

Fast BFS-Based Triangle Counting on GPUs

In this paper, we propose a novel method to compute triangle counting on GPUs. Unlike previous formulations of graph matching, our approach is BFS-based by traversing the graph in an all-source-BFS manner and thus can be mapped onto GPUs in a massively parallel fashion. Our implementation uses the Gunrock programming model and we evaluate our implementation in runtime and memory consumption compared with previous state-of-the-art work. We sustain a peak traversed-edges-per-second (TEPS) rate of nearly 10 GTEPS. Our algorithm is the most scalable and parallel among all existing GPU implementations and also outperforms all existing CPU distributed implementations. This work specifically focuses on leveraging our implementation on the triangle counting problem for the Subgraph Isomorphism Graph Challenge 2019, demonstrating a geometric mean speedup over the 2018 champion of 3.84x.

cs.DC↗

arXiv2018

Implementing Push-Pull Efficiently in GraphBLAS

We factor Beamer's push-pull, also known as direction-optimized breadth-first-search (DOBFS) into 3 separable optimizations, and analyze them for generalizability, asymptotic speedup, and contribution to overall speedup. We demonstrate that masking is critical for high performance and can be generalized to all graph algorithms where the sparsity pattern of the output is known a priori. We show that these graph algorithm optimizations, which together constitute DOBFS, can be neatly and separably described using linear algebra and can be expressed in the GraphBLAS linear-algebra-based framework. We provide experimental evidence that with these optimizations, a DOBFS expressed in a linear-algebra-based graph framework attains competitive performance with state-of-the-art graph frameworks on the GPU and on a multi-threaded CPU, achieving 101 GTEPS on a Scale 22 RMAT graph.

cs.DC↗

arXiv2018

Design Principles for Sparse Matrix Multiplication on the GPU

We implement two novel algorithms for sparse-matrix dense-matrix multiplication (SpMM) on the GPU. Our algorithms expect the sparse input in the popular compressed-sparse-row (CSR) format and thus do not require expensive format conversion. While previous SpMM work concentrates on thread-level parallelism, we additionally focus on latency hiding with instruction-level parallelism and load-balancing. We show, both theoretically and experimentally, that the proposed SpMM is a better fit for the GPU than previous approaches. We identify a key memory access pattern that allows efficient access into both input and output matrices that is crucial to getting excellent performance on SpMM. By combining these two ingredients---(i) merge-based load-balancing and (ii) row-major coalesced memory access---we demonstrate a 4.1x peak speedup and a 31.7% geomean speedup over state-of-the-art SpMM implementations on real-world datasets.

cs.DC↗

arXiv2018

A Comparative Study on Exact Triangle Counting Algorithms on the GPU

We implement exact triangle counting in graphs on the GPU using three different methodologies: subgraph matching to a triangle pattern; programmable graph analytics, with a set-intersection approach; and a matrix formulation based on sparse matrix-matrix multiplies. All three deliver best-of-class performance over CPU implementations and over comparable GPU implementations, with the graph-analytic approach achieving the best performance due to its ability to exploit efficient filtering steps to remove unnecessary work and its high-performance set-intersection core.

cs.DC↗

arXiv2018

Sampling Conditions for Conforming Voronoi Meshing by the VoroCrust Algorithm

We study the problem of decomposing a volume bounded by a smooth surface into a collection of Voronoi cells. Unlike the dual problem of conforming Delaunay meshing, a principled solution to this problem for generic smooth surfaces remained elusive. VoroCrust leverages ideas from $α$-shapes and the power crust algorithm to produce unweighted Voronoi cells conforming to the surface, yielding the first provably-correct algorithm for this problem. Given an $ε$-sample on the bounding surface, with a weak $σ$-sparsity condition, we work with the balls of radius $δ$ times the local feature size centered at each sample. The corners of this union of balls are the Voronoi sites, on both sides of the surface. The facets common to cells on opposite sides reconstruct the surface. For appropriate values of $ε$, $σ$ and $δ$, we prove that the surface reconstruction is isotopic to the bounding surface. With the surface protected, the enclosed volume can be further decomposed into an isotopic volume mesh of fat Voronoi cells by generating a bounded number of sites in its interior. Compared to state-of-the-art methods based on clipping, VoroCrust cells are full Voronoi cells, with convexity and fatness guarantees. Compared to the power crust algorithm, VoroCrust cells are not filtered, are unweighted, and offer greater flexibility in meshing the enclosed volume by either structured grids or random samples.

cs.CG↗

arXiv2018

Scalable Breadth-First Search on a GPU Cluster

On a GPU cluster, the ratio of high computing power to communication bandwidth makes scaling breadth-first search (BFS) on a scale-free graph extremely challenging. By separating high and low out-degree vertices, we present an implementation with scalable computation and a model for scalable communication for BFS and direction-optimized BFS. Our communication model uses global reduction for high-degree vertices, and point-to-point transmission for low-degree vertices. Leveraging the characteristics of degree separation, we reduce the graph size to one third of the conventional edge list representation. With several other optimizations, we observe linear weak scaling as we increase the number of GPUs, and achieve 259.8 GTEPS on a scale-33 Graph500 RMAT graph with 124 GPUs on the latest CORAL early access system.

cs.DC↗

arXiv2018

GPU LSM: A Dynamic Dictionary Data Structure for the GPU

We develop a dynamic dictionary data structure for the GPU, supporting fast insertions and deletions, based on the Log Structured Merge tree (LSM). Our implementation on an NVIDIA K40c GPU has an average update (insertion or deletion) rate of 225 M elements/s, 13.5x faster than merging items into a sorted array. The GPU LSM supports the retrieval operations of lookup, count, and range query operations with an average rate of 75 M, 32 M and 23 M queries/s respectively. The trade-off for the dynamic updates is that the sorted array is almost twice as fast on retrievals. We believe that our GPU LSM is the first dynamic general-purpose dictionary data structure for the GPU.

cs.DC↗

arXiv2018

A Dynamic Hash Table for the GPU

We design and implement a fully concurrent dynamic hash table for GPUs with comparable performance to the state of the art static hash tables. We propose a warp-cooperative work sharing strategy that reduces branch divergence and provides an efficient alternative to the traditional way of per-thread (or per-warp) work assignment and processing. By using this strategy, we build a dynamic non-blocking concurrent linked list, the slab list, that supports asynchronous, concurrent updates (insertions and deletions) as well as search queries. We use the slab list to implement a dynamic hash table with chaining (the slab hash). On an NVIDIA Tesla K40c GPU, the slab hash performs updates with up to 512 M updates/s and processes search queries with up to 937 M queries/s. We also design a warp-synchronous dynamic memory allocator, SlabAlloc, that suits the high performance needs of the slab hash. SlabAlloc dynamically allocates memory at a rate of 600 M allocations/s, which is up to 37x faster than alternative methods in similar scenarios.

cs.DC↗

arXiv2017

GPU Multisplit: an extended study of a parallel algorithm

Multisplit is a broadly useful parallel primitive that permutes its input data into contiguous buckets or bins, where the function that categorizes an element into a bucket is provided by the programmer. Due to the lack of an efficient multisplit on GPUs, programmers often choose to implement multisplit with a sort. One way is to first generate an auxiliary array of bucket IDs and then sort input data based on it. In case smaller indexed buckets possess smaller valued keys, another way for multisplit is to directly sort input data. Both methods are inefficient and require more work than necessary: the former requires more expensive data movements while the latter spends unnecessary effort in sorting elements within each bucket. In this work, we provide a parallel model and multiple implementations for the multisplit problem. Our principal focus is multisplit for a small (up to 256) number of buckets. We use warp-synchronous programming models and emphasize warp-wide communications to avoid branch divergence and reduce memory usage. We also hierarchically reorder input elements to achieve better coalescing of global memory accesses. On a GeForce GTX 1080 GPU, we can reach a peak throughput of 18.93 Gkeys/s (or 11.68 Gpairs/s) for a key-only (or key-value) multisplit. Finally, we demonstrate how multisplit can be used as a building block for radix sort. In our multisplit-based sort implementation, we achieve comparable performance to the fastest GPU sort routines, sorting 32-bit keys (and key-value pairs) with a throughput of 3.0 G keys/s (and 2.1 Gpair/s).

cs.DC↗

arXiv2017

Multi-GPU Graph Analytics

We present a single-node, multi-GPU programmable graph processing library that allows programmers to easily extend single-GPU graph algorithms to achieve scalable performance on large graphs with billions of edges. Directly using the single-GPU implementations, our design only requires programmers to specify a few algorithm-dependent concerns, hiding most multi-GPU related implementation details. We analyze the theoretical and practical limits to scalability in the context of varying graph primitives and datasets. We describe several optimizations, such as direction optimizing traversal, and a just-enough memory allocation scheme, for better performance and smaller memory consumption. Compared to previous work, we achieve best-of-class performance across operations and datasets, including excellent strong and weak scalability on most primitives as we increase the number of GPUs in the system.

cs.DC↗

arXiv2017

Gunrock: GPU Graph Analytics

For large-scale graph analytics on the GPU, the irregularity of data access and control flow, and the complexity of programming GPUs, have presented two significant challenges to developing a programmable high-performance graph library. "Gunrock", our graph-processing system designed specifically for the GPU, uses a high-level, bulk-synchronous, data-centric abstraction focused on operations on a vertex or edge frontier. Gunrock achieves a balance between performance and expressiveness by coupling high performance GPU computing primitives and optimization strategies with a high-level programming model that allows programmers to quickly develop new graph primitives with small code size and minimal GPU programming knowledge. We characterize the performance of various optimization strategies and evaluate Gunrock's overall performance on different GPU architectures on a wide range of graph primitives that span from traversal-based algorithms and ranking algorithms, to triangle counting and bipartite-graph-based algorithms. The results show that on a single GPU, Gunrock has on average at least an order of magnitude speedup over Boost and PowerGraph, comparable performance to the fastest GPU hardwired primitives and CPU shared-memory graph libraries such as Ligra and Galois, and better performance than any other GPU high-level graph library.

cs.DC↗

arXiv2016

Mathematical Foundations of the GraphBLAS

The GraphBLAS standard (GraphBlas.org) is being developed to bring the potential of matrix based graph algorithms to the broadest possible audience. Mathematically the Graph- BLAS defines a core set of matrix-based graph operations that can be used to implement a wide class of graph algorithms in a wide range of programming environments. This paper provides an introduction to the mathematics of the GraphBLAS. Graphs represent connections between vertices with edges. Matrices can represent a wide range of graphs using adjacency matrices or incidence matrices. Adjacency matrices are often easier to analyze while incidence matrices are often better for representing data. Fortunately, the two are easily connected by matrix mul- tiplication. A key feature of matrix mathematics is that a very small number of matrix operations can be used to manipulate a very wide range of graphs. This composability of small number of operations is the foundation of the GraphBLAS. A standard such as the GraphBLAS can only be effective if it has low performance overhead. Performance measurements of prototype GraphBLAS implementations indicate that the overhead is low.

cs.MS↗

arXiv2016

Gunrock: A High-Performance Graph Processing Library on the GPU

For large-scale graph analytics on the GPU, the irregularity of data access and control flow, and the complexity of programming GPUs have been two significant challenges for developing a programmable high-performance graph library. "Gunrock", our graph-processing system designed specifically for the GPU, uses a high-level, bulk-synchronous, data-centric abstraction focused on operations on a vertex or edge frontier. Gunrock achieves a balance between performance and expressiveness by coupling high performance GPU computing primitives and optimization strategies with a high-level programming model that allows programmers to quickly develop new graph primitives with small code size and minimal GPU programming knowledge. We evaluate Gunrock on five key graph primitives and show that Gunrock has on average at least an order of magnitude speedup over Boost and PowerGraph, comparable performance to the fastest GPU hardwired primitives, and better performance than any other GPU high-level graph library.

cs.DC↗