The new reality of exascale computing faces many challenges in achieving optimal performance on large numbers of nodes. A key challenge is the efficient utilization of the message-passing interface (MPI), a critical component for process communication. This paper explores communication optimization strategies to harness the GPU-accelerated architectures of these supercomputers. We focus on MPI applications where processors form a two-dimensional process grid, a common arrangement in applications involving dense matrix operations. This configuration offers a unique opportunity to implement innovative strategies to improve performance and maintain effective load distribution. We study two applications— Dist-FW (Apsp:all-pair-shortest-path) and HPL-MxP (LU factorization with Mixed precision)—on two accelerated systems: Summit (IBM Power and NVIDIA V100) and Frontier (AMD EPYC and MI250X). These supercomputers are operated by the Oak Ridge Leadership Computing Facility (OLCF) and are currently ranked #1 and #5 on the Top500 list. We show how to scale up both applications to exascale levels and tackle the MPI challenges related to implementation, synchronization, and performance. We also compare the performance of several communication strategies at an unprecedented scale. Accurately predicting application performance becomes crucial for cost reduction as the computational scale grows. To address this, we suggest a hyperbolic model as a better alternative to the traditional one-sided asymptotic model for predicting future application performance at such large scales.

We are motivated by newly proposed methods for mining large-scale corpora of scholarly publications (e.g., full biomedical literature), which consists of tens of millions of papers spanning decades of research. In this setting, analysts seek to discover relationships among concepts. They construct graph representations from annotated text databases and then formulate the relationship-mining problem as an all-pairs shortest paths (APSP) and validate connective paths against curated biomedical knowledge graphs (e.g., Spoke). In this context, we present Coast (Exascale Communication-Optimized All-Pairs Shortest Path) and demonstrate 1.004 EF/s on 9,200 Frontier nodes (73,600 GCDs). We develop hyperbolic performance models (HYPERMOD), which guide optimizations and parametric tuning. The proposed Coast algorithm achieved the memory constant parallel efficiency of 99% in the single-precision tropical semiring. Looking forward, Coast will enable the integration of scholarly corpora like PubMed into the Spoke biomedical knowledge graph.

This study aims to effectively analyze and visualize the concept to concept network derived from the COVID-19 Open Research Dataset (CORD-19) dataset, where we have more than 48,000 concepts with more than 300,000 relationships between concepts. In analyzing networks, we focus on finding relationship patterns between the coronavirus disease 2019 (COVID-19) concepts and other concepts. Given the node and edge datasets, we construct directional graphs and calculate all pair shortest paths based on multiple edge weight schemes. However, statistical metrics are not sufficient to identify specific relationships represented in the network. Therefore, we also propose a visual analytics approach to effectively understand the knowledge graph. Our highly interactive visual analytics allows users to effectively analyze the evolving graphs and (COVID-19) concept nodes and other nodes related to the COVID-19 nodes. We envision that this study will pave the path to develop strategies to provide more accurate and scalable predictive analysis on knowledge graphs related to CORD19 and other biomedical knowledge graphs.

We present an optimized Floyd-Warshall (Floyd-Warshall) algorithm that computes the All-pairs shortest path (APSP) for GPU accelerated clusters. The Floyd-Warshall algorithm due to its structural similarities to matrix-multiplication is well suited for highly parallel GPU architectures. To achieve high parallel efficiency, we address two key algorithmic challenges: reducing high communication overhead and addressing limited GPU memory. To reduce high communication costs, we redesign the parallel (a) to expose more parallelism, (b) aggressively overlap communication and computation with pipelined and asynchronous scheduling of operations, and (c) tailored MPI-collective. To cope with limited GPU memory, we employ an offload model, where the data resides on the host and is transferred to GPU on-demand. The proposed optimizations are supported with detailed performance models for tuning. Our optimized parallel Floyd-Warshall implementation is up to 5x faster than a strong baseline and achieves 8.1 PetaFLOPS/sec on 256~nodes of the Summit supercomputer at Oak Ridge National Laboratory. This performance represents 70% of the theoretical peak and 80% parallel efficiency. The offload algorithm can handle 2.5x larger graphs with a 20% increase in overall running time.

Training neural networks for neuromorphic deployment is non-trivial. There have been a variety of approaches proposed to adapt back-propagation or back-propagation-like algorithms appropriate for training. Considering that these networks often have very different performance characteristics than traditional neural networks, it is often unclear how to set either the network topology or the hyperparameters to achieve optimal performance. In this work, we introduce a Bayesian approach for optimizing the hyperparameters of an algorithm for training binary communication networks that can be deployed to neuromorphic hardware. We show that by optimizing the hyperparameters on this algorithm for each dataset, we can achieve improvements in accuracy over the previous state-of-the-art for this algorithm on each dataset (by up to 15 percent). This jump in performance continues to emphasize the potential when converting traditional neural networks to binary communication applicable to neuromorphic hardware.

Designing spiking neural networks for neuromorphic deployment is a non-trivial task. It is further complicated when there are resource constraints for the neuromorphic implementation, such as size or power constraints, that may be present in edge applications. In this work, we utilize a previously presented approach, EONS, to design spiking neural networks for a memristive neuromorphic implementation for scientific data applications. We specifically use a multi-objective approach in EONS to maximize network accuracy on the scientific data application task, but also to minimize network size and energy. We illustrate that EONS determines both the network structure and the parameters, removing the burden from the user on determining the appropriate spiking neural network structure, and we show that the resulting networks are very different from the layered structure of typical neural networks. Finally, we show that the multi-objective approach produces smaller, more energy efficient networks than the original EONS approach and produces comparable accuracy to a back-propagation style training approach.

We are motivated by newly proposed methods for data mining large-scale corpora of scholarly publications, such as the full biomedical literature, which may consist of tens of millions of papers spanning decades of research. In this setting, analysts seek to discover how concepts relate to one another. They construct graph representations from annotated text databases and then formulate the relationship-mining problem as one of computing all-pairs shortest paths (APSP), which becomes a significant bottleneck. In this context, we present a new high-performance algorithm and implementation of the Floyd-Warshall algorithm for distributed-memory parallel computers accelerated by GPUs, which we call DSNAPSHOT (Distributed Accelerated Semiring All-Pairs Shortest Path). For our largest experiments, we ran DSNAPSHOT on a connected input graph with millions of vertices using 4, 096nodes (24,576GPUs) of the Oak Ridge National Laboratory's Summit supercomputer system. We find DSNAPSHOT achieves a sustained performance of $136\times 10^{15}$ floating-point operations per second (136petaflop/s) at a parallel efficiency of 90% under weak scaling and, in absolute speed, 70% of the best possible performance given our computation (in the single-precision tropical semiring or “min-plus” algebra). Looking forward, we believe this novel capability will enable the mining of scholarly knowledge corpora when embedded and integrated into artificial intelligence-driven natural language processing workflows at scale.

Deep learning is actively used in a wide range of fields for scientific discovery. To effectively apply deep learning to a particular problem, it is important to select an appropriate network architecture and other hyper-parameters (at each layer). Evolving architectures and hyper-parameters using a genetic algorithm is one current approach to search the huge space of all possible configurations to find those more optimal for the problem. However, examining an evolutionary process and tuning the genetic algorithm are challenging, pushing most users to treat the process as a black box. To address this challenge, we propose a visualization system for evolutionary neural networks for deep learning. The key feature of our visualization system is to provide a visual analytics environment for evaluating a genetic algorithm in order to improve the underlying operations to reduce time to find good solutions. Our system is able to not only visualize how a genetic algorithm traverses its search space but also allows users to examine evolving networks in-depth to get insights to improve performance through interactive visualization components.

Designing a neuromorphic computing system involves selection of several hyperparameters that not only affect the accuracy of the framework, but also the energy efficiency and speed of inference and training. These hyperparameters might be inherent to the training of the spiking neural network (SNN), the input/output encoding of the real-world data to spikes, or the underlying neuromorphic hardware. In this work, we present a Bayesian-based hyperparameter optimization approach for spiking neuromorphic systems, and we show how this optimization framework can lead to significant improvement in designing accurate neuromorphic computing systems. In particular, we show that this hyperparameter optimization approach can discover the same optimal hyperparameter set for input encoding as a grid search, but with far fewer evaluations and far less time. We also show the impact of hardware-specific hyperparameters on the performance of the system, and we demonstrate that by optimizing these hyperparameters, we can achieve significantly better application performance.

Deep learning, through the use of neural networks, has demonstrated remarkable ability to automate many routine tasks when presented with sufficient data for training. The neural network architecture (e.g. number of layers, types of layers, connections between layers, etc.) plays a critical role in determining what, if anything, the neural network is able to learn from the training data. The trend for neural network architectures, especially those trained on ImageNet, has been to grow ever deeper and more complex. The result has been ever increasing accuracy on benchmark datasets with the cost of increased computational demands. In this paper we demonstrate that neural network architectures can be automatically generated, tailored for a specific application, with dual objectives: accuracy of prediction and speed of prediction. Using MENNDL- an HPC-enabled software stack for neural architecture search-we generate a neural network with comparable accuracy to state-of-the-art networks on a cancer pathology dataset that is also 16× faster at inference. The speedup in inference is necessary because of the volume and velocity of cancer pathology data; specifically, the previous state-of-the-art networks are too slow for individual researchers without access to HPC systems to keep pace with the rate of data generation. Our new model enables researchers with modest computational resources to analyze newly generated data faster than it is collected.

As deep neural networks have been deployed in more and more applications over the past half decade and are finding their way into an ever increasing number of operational systems, their energy consumption becomes a concern whether running in the datacenter or on edge devices. Hyperparameter optimization and automated network design for deep learning is a quickly growing field, but much of the focus has remained only on optimizing for the performance of the machine learning task. In this work, we demonstrate that the best performing networks created through this automated network design process have radically different computational characteristics (e.g. energy usage, model size, inference time), presenting the opportunity to utilize this optimization process to make deep learning networks more energy efficient and deployable to smaller devices. Optimizing for these computational characteristics is critical as the number of applications of deep learning continues to expand.

We present a deep learning approach for vertex reconstruction of neutrino-nucleus interaction events, a problem in the domain of high energy physics. In this approach, we combine both energy and timing data that are collected in the MIN-ERvA detector to perform classification and regression tasks. We show that the resulting network achieves higher accuracy than previous results while requiring a smaller model size and less training time. In particular, the proposed model outperforms the state-of-the-art by 4.00% on classification accuracy. For the regression task, our model achieves 0.9919 on the coefficient of determination, higher than the previous work (0.96).

Current deep learning approaches have been very successful using convolutional neural networks trained on large graphical-processing-unit-based computers. Three limitations of this approach are that (1) they are based on a simple layered network topology, i.e., highly connected layers, without intra-layer connections; (2) the networks are manually configured to achieve optimal results, and (3) the implementation of the network model is expensive in both cost and power. In this article, we evaluate deep learning models using three different computing architectures to address these problems: quantum computing to train complex topologies, high performance computing to automatically determine network topology, and neuromorphic computing for a low-power hardware implementation. We use the MNIST dataset for our experiment, due to input size limitations of current quantum computers. Our results show the feasibility of using the three architectures in tandem to address the above deep learning limitations. We show that a quantum computer can find high quality values of intra-layer connection weights in a tractable time as the complexity of the network increases, a high performance computer can find optimal layer-based topologies, and a neuromorphic computer can represent the complex topology and weights derived from the other architectures in low power memristive hardware.

No abstract available.

Enterprise networks are comprised of thousands of interconnected computer hosts, each of which is capable of creating, removing, and exchanging data according to the needs of their users. Thus, the distribution of high-value, sensitive, and proprietary information across enterprise networks is poorly managed and understood. A significant technology gap in information security is the inability to automatically quantify the value of the information contained on each host in a network. Such insight would allow an enterprise to scale its defenses, react intelligently to an intrusion, manage its configuration audits, and understand the leak potential in the event that a host is compromised. This paper outlines a novel approach to the automated determination of the value of the information contained on a host computer. It involves the classification of each text document on the host machine using the frequency of the document's terms and phrases. A host information value is computed using an enterprise-defined weighting schema and applying it to a host's document distribution. The method is adaptable to specific organizational information needs, requires manual intervention only during schema creation, and is repeatable and consistent regardless of changes in information on the host machines.

Abstract—Most commercial intrusion detections systems (IDS) can produce a very high volume of alerts, and are typically plagued by a high false positive rate. The approach described here uses Splunk to aggregate IDS alerts. The aggregated IDS alerts are retrieved from Splunk programmatically and are then clustered using text analysis and visualized using a sunburst diagram to provide an additional understanding of the data. The equivalent of what the cluster analysis and visualization provides would require numerous detailed queries using Splunk and considerable manual effort. Keywords-hierarchical clustering, sunburst visualization, IDS analysis, vector space model I.

Document clustering plays an important role in data mining systems. Recently, a flocking-based document clustering algorithm has been proposed to solve the problem through simulation resembling the flocking behavior of birds in nature. This method is superior to other clustering algorithms, including k-means, in the sense that the outcome is not sensitive to the initial state. One limitation of this approach is that the algorithmic complexity is inherently quadratic in the number of documents. As a result, execution time becomes a bottleneck with large number of documents. In this paper, we assess the benefits of exploiting the computational power of Beowulf-like clusters equipped with contemporary Graphics Processing Units (GPUs) as a means to significantly reduce the runtime of flocking-based document clustering. Our framework scales up to over one million documents processed simultaneously in a sixteen-node moderate GPU cluster. Results are also compared to a four-node cluster with higher-end GPUs. On these clusters, we observe 30X-50X speedups, which demonstrate the potential of GPU clusters to efficiently solve massive data mining problems. Such speedups combined with the scalability potential and accelerator-based parallelization are unique in the domain of document-based data mining, to the best of our knowledge.

Assessing the potential property and social impacts of an event, such as tornado or wildfire, continues to be a challenging research area. From financial markets to disaster management to epidemiology, the importance of understanding the impacts that events create cannot be understated. Our work describes an approach to fuse information from multiple sources, then to analyze the information cycles to identify prior temporal patterns related to the impact of an event. This approach is then applied to the analysis of news reports from multiple news sources pertaining to several different natural disasters. Results show that our approach can project the severity of the impacts of certain natural disasters, such as heat waves on droughts and wild fires. In addition, results show that specific types of disaster consistently produce similar impacts when each time they occur.

The Particle Swarm Optimization (PSO) clustering algorithm can generate more compact clustering results than the traditional K-means clustering algorithm. However, when clustering high dimensional datasets, the PSO clustering algorithm is notoriously slow because its computation cost increases exponentially with the size of the dataset dimension. Dimensionality reduction techniques offer solutions that both significantly improve the computation time, and yield reasonably accurate clustering results in high dimensional data analysis. In this paper, we introduce research that combines different dimensionality reduction techniques with the PSO clustering algorithm in order to reduce the complexity of high dimensional datasets and speed up the PSO clustering process. We report significant improvements in total runtime. Moreover, the clustering accuracy of the dimensionality reduction PSO clustering algorithm is comparable to the one that uses full dimension space.

Particle swarm optimization (PSO) is a population-based stochastic optimization technique, which can be used to find an optimal, or near optimal, solution to a numerical and qualitative problem. In PSO algorithm, the problem solution emerges from the interactions among many simple individual agents called particles. In the real world, we have to frequently deal with searching and tracking an optimal solution in a dynamical and noisy environment. This demands that the algorithm not only find the optimal solution but also track the trajectory of the non-stationary solution. The traditional PSO algorithm lacks the ability to track the changing optimal solution in a dynamic and noisy environment. In this paper, we present a distributed adaptive PSO (DAPSO) algorithm that can be used to track a non-stationary optimal solution in a dynamically changing and noisy environment.

To better understand insurgent activities and asymmetric warfare, a social adaptive model for modeling multiple insurgent groups attacking multiple military and civilian targets is proposed and investigated. This report presents a pilot study using the particle swarm modeling, a widely used non-linear optimal tool, to model the emergence of insurgency campaign. The objective of this research is to apply the particle swarm metaphor as a model of insurgent social adaptation for the dynamically changing environment and to provide insight and understanding of insurgency warfare. Our results show that unified leadership, strategic planning, and effective communication between insurgent groups are not the necessary requirements for insurgents to efficiently attain their objectives.

There exists an enormous amount of information available via the Internet. Much of this data is in the form of text-based documents. These documents cover a variety of topics that are vitally important to the scientific, business, and defense/security communities. Currently, there are a many techniques for processing and analyzing such data. However, the ability to quickly characterize a large set of documents still proves challenging. Previous work has successfully demonstrated the use of a genetic algorithm for providing a representative subset for text documents via adaptive sampling. In this work, we further expand and explore this approach on much larger data sets using a parallel Genetic Algorithm (GA) with adaptive parameter control. Experimental results are presented and discussed.

In this paper, we propose a new term weighting scheme called term frequency-inverse corpus frequency (TF-ICF). It does not require term frequency information from other documents within the document collection and thus, it enables us to generate the document vectors of N streaming documents in linear time. In the context of a machine learning application, unsupervised document clustering, we evaluated the effectiveness of the proposed approach in comparison to five widely used term weighting schemes through extensive experimentation. Our results show that TF-ICF can produce document clusters that are of comparable quality as those generated by the widely recognized term weighting schemes and it is significantly faster than those methods

Intelligence analysts are currently overwhelmed with the amount of information streams generated everyday. There is a lack of comprehensive tool that can real-time analyze the information streams. Document clustering analysis plays an important role in improving the accuracy of information retrieval. However, most clustering technologies can only be applied for analyzing the static document collection because they normally require a large amount of computation resource and long time to get accurate result. It is very difficult to cluster a dynamic changed text information streams on an individual computer. Our early research has resulted in a dynamic reactive flock clustering algorithm which can continually refine the clustering result and quickly react to the change of document contents. This character makes the algorithm suitable for cluster analyzing dynamic changed document information, such as text information stream. Because of the decentralized character of this algorithm, a distributed approach is a very natural way to increase the clustering speed of the algorithm. In this paper, we present a distributed multi-agent flocking approach for the text information stream clustering and discuss the decentralized architectures and communication schemes for load balance and status information synchronization in this approach.

Recent years saw the rapid development of peer-to-peer (P2P) networks in a great variety of applications. However, similarity-based k-nearest-neighbor retrieval (k-NN) is still a challenging task in P2P networks due to the multiple constraints such as the dynamic topologies and the unpredictable data updates. Caching is an attractive solution that reduces network traffic and hence could remedy the technological constraints of P2P networks. However, traditional caching techniques have some major shortcomings that make them unsuitable for similarity search, such as the lack of semantic locality representation and the rigidness of exact matching on data objects. To facilitate the efficient similarity search, we propose semantic-aware caching scheme (SAC) in this paper. The proposed scheme is hierarchy-free, fully dynamic, non-flooding, and do not add much system overhead. By exploring the content distribution, SAC drastically reduces the cost of similarity-based k-NN retrieval in P2P networks. The performance of SAC is evaluated through simulation study and compared against several search schemes as advanced in the literature.

In the real world, we have to frequently deal with searching for and tracking an optimal solution in a dynamic environment. This demands that the algorithm not only find the optimal solution but also track the trajectory of the solution in a dynamic environment. Particle swarm optimization (PSO) is a population-based stochastic optimization technique, which can find an optimal, or near optimal, solution to a numerical and qualitative problem. However, the traditional PSO algorithm lacks the ability to track the optimal solution in a dynamic environment. In this paper, we present a modified PSO algorithm that can be used for tracking a non-stationary optimal solution in a dynamically changing environment.

Fast and high-quality document clustering algorithms play an important role in effectively navigating, summarizing, and organizing information. Recent studies have shown that partitional clustering algorithms are more suitable for clustering large datasets. However, the K-means algorithm, the most commonly used partitional clustering algorithm, can only generate a local optimal solution. In this paper, we present a particle swarm optimization (PSO) document clustering algorithm. Contrary to the localized searching of the K-means algorithm, the PSO clustering algorithm performs a globalized search in the entire solution space. In the experiments we conducted, we applied the PSO, K-means and hybrid PSO clustering algorithm on four different text document datasets. The number of documents in the datasets ranges from 204 to over 800, and the number of terms ranges from over 5000 to over 7000. The results illustrate that the hybrid PSO algorithm can generate more compact clustering results than the K-means algorithm.

The introduction of object-oriented technology does not appear to hinder overall productivity on new large commercial projects, but nor does it seem to improve it in the first two product generations. In practice, the governing influence may be the business workflow, and not the methodology. Copyright © 1999 John Wiley & Sons, Ltd.