Search NASA⌕ Search

SEARCH · Search NASA

Results for “VECTOR PROCESSING”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 865 records · Page 48

Hadron spectroscopy of pseudoscalar and vector meson photoproduction using linearly polarised photons at CLAS

The field of hadron spectroscopy is informed by experiments which study excited states of the nucleon by detecting the decay products following excitation with a hadronic, leptonic or electromagnetic probe. Experiments using a polarised beam of photons provide an important contribution to the world data. This thesis describes the analysis of data from one such experiment, the CEBAF Large Acceptance Spectrometer at Jefferson Laboratory, Virginia, USA. In the experiment, an incident beam of electrons passes through a diamond radiator producing a linearly polarised photon beam of energy from 1.1 to 2.1 GeV on a liquid hydrogen target. The analysis method and results for two reaction channels are presented. In the first analysis, the polarisation observables {?, P, T, Ox, Oz} for the reaction ??? p ? K0 S ?+ are extracted simultaneously using likelihood sampling with Markov-Chain Monte-Carlo. The results are presented as angular distributions (cos ?K0S) for bins in photon energy E? for a total of 21 bins. The values extracted for T , Ox and Oz are a first measurement of these quantities for this reaction and the beam asymmetry (?) measurement supplements the one previous measurement for this quantity and extends the energy range over which it has been extracted. The measurement of the recoil polarisation, P, is consistent with previous measurements. In a second analysis, the spin density matrix elements (SDMEs) for ? p ? p ? are extracted using similar analysis techniques. The results are presented as angular distributions (cos ?? ) for bins in photon energy E? for a total of 24 bins. This work is the first to extract all nine SDMEs for the full angular range of the ? meson production. A comparison is made to the predictions of the Vector Meson Dominance (VMD) model and the results provide evidence for contributing processes other than VMD, particularly at more backward angles of the ? meson production. The data also indicate non-helicity conserving process in both the s-channel and the t-channel. The results from the first analysis have been passed to the Juelich-Bonn theory group and the effect of including the new data within their dynamical coupled channel model is described. Several nucleon resonances are affected and the impact on the pole positions, widths and photon decay amplitudes of the affected resonances is described.

Clark, Louise↗

Cryo-electron Microscopy of Adeno-associated Virus

Adeno-associated virus (AAV) has a single-stranded DNA genome encapsidated in a small icosahedrally symmetric protein shell with 60 subunits. AAV is the leading delivery vector in emerging gene therapy treatments for inherited disorders, so its structure and molecular interactions with human hosts are of intense interest. A wide array of electron microscopic approaches have been used to visualize the virus and its complexes, depending on the scientific question, technology available, and amenability of the sample. Approaches range from subvolume tomographic analyses of complexes with large and flexible host proteins to detailed analysis of atomic interactions within the virus and with small ligands at resolutions as high as 1.6 Å. Analyses have led to the reclassification of glycan receptors as attachment factors, to structures with a new-found receptor protein, to identification of the epitopes of antibodies, and a new understanding of possible neutralization mechanisms. AAV is now well-enough characterized that it has also become a model system for EM methods development. Heralding a new era, cryo-EM is now also being deployed as an analytic tool in the process development and production quality control of high value pharmaceutical biologics, namely AAV vectors.

59 BASIC BIOLOGICAL SCIENCES↗

Data-driven predictive probability density function control of fiber length stochastic distribution shaping in refining process

Pulp is the most important raw material in paper in paper industries, whose Fiber Length Stochastic Distribution (FLSD) shaping directly determines the energy consumption and paper quality of the subsequent papermaking processes. However, the mean and variance are insufficient to describe the output FLSD shaping, which display non-Gaussian distribution properties. Therefore, the traditional control, method based on the mean and variance of fiber length is difficult to control the FLSD shaping effectively. In this paper, a novel data-driven predictive probability density function (PDF) control method is proposed for output FLSD shaping in refining process. Primarily, in order to improve the approximation accuracy of the PDF employing RBF neural network, the parameters (center value and width) of each RBF basis functions are tuned via utilizing iterative learning control (ILC) law, and the corresponding estimations of weights law can be obtained. Secondly, considering that the conventional linear model of weights vector has drawbacks of low accuracy and weak generalization ability, random vector functional link networks (RVFLNs) based data driven nonlinear modelling method is employed to characterize the prediction model between the input variables and weights vector. Finally, in order to reduce the randomness of the output FLSD, a minimum entropy control method under mean constraint is employed to design the predictive PDF controller for the output FLSD shaping with the help of the established nonlinear stochastic distribution model. Both simulations and experiments are given to illustrate the effectiveness and practicability of the proposed method.

Probability debsity function, Fiber length distrib↗

Study of direct photon production with heavy flavor jets in $p\overline{p}$ collisions at $\sqrt{s}=1.96$ TeV with DZero detector

In hadronic collisions, photons ($\gamma$) with high energies emerge unaltered from the hard parton-parton interaction and therefore provide a clean probe of the underlying hard-scattering dynamics. Photons produced in these interactions (called direct or prompt) in association with one or more bottom ($b$)-quark jets provide an important test of perturbative Quantum Chromodynamics (QCD) predictions at large hard-scattering scales $Q$ and over a wide range of parton momentum fractions. In addition, the study of these processes also provides information about the parton density functions of $b$ quarks and gluons ($g$), which still have substantial uncertainties. In $p\bar{p}$ collisions, \gb-jet events are produced primarily through the Compton process $gb\to \gamma b$, which dominates for low and moderate photon transverse momenta ($\Ptg$), and through quark-antiquark anni hilation followed by $g \to b\bar{b}$ gluon splitting $q\bar{q}\to \gamma g \to \gamma b\bar{b}$, which dominates at high $\Ptg$. The final state with $b$-quark pair production, $p\bar{p} \rightarrow \gamma+b\bar{b}$, is mainly produced via $q\bar{q}\to \gamma b\bar{b}$ and $gg\to \gamma b\bar{b}$ scatterings. The $\gamma+2~b$-jet process is a crucial component of background in measurements of, for example, $t\bar{t} \gamma$ coupling and in some searches for new phenomena.\\ This thesis presents the first measurements of the differential cross section ${\rm d}\sigma/{\rm d}\Ptg$ for the production of an isolated photon in association with at least two $b$-quark jets. The ratio of differential production cross sections for $\gamma+2~b$-jets to $\gamma+b$-jet as a function of \ptg is also presented. The measurement of the ratio of cross sections leads to cancellation of various experimental and theoretical uncertainties, allowing a more precise comparison with the theoretical predictions. The results are based on the proton-antiproton ! collision data corresponding to an integrated luminosity of $8.7$~fb$^{-1}$ at $\sqrt{s}=$1.96~\TeV collected with the \DO detector at the Fermilab Tevatron Collider. %In this measurement, we follow an inclusive approach by allowing the final state with any additional jet(s) on top of the studied $b$-quark jets. Inclusive $\gamma+2~b$-jet production may also originate from partonic subprocesses involving parton fragmentation into a photon. However, photon isolation requirements significantly reduces the contributions from such processes. The large data sample and use of advanced photon and $b$-jet identification tools enable us to measure the $\gamma+2~b$-jet production cross section differentially as a function of $\Ptg$. This allows for probing the dynamics of the production process over a wide kinematic range not studied before in other measurements of a vector boson + $b$-jet final state. The measured cross sections and their ratios are compared to the NLO perturbative Q CD calculations as well as predictions based on the $k_{\rm T}$-factorization approach and those from the {\sc sherpa} and {\sc pythia} Monte Carlo event generators. We also measure the ratio of cross sections, $\sigma(p\bar{p}\rightarrow Z+2~b~\text{jets})$/$\sigma(p\bar{p}\rightarrow Z+\text{2~jets})$, for associated production of a $Z$ boson with at least two jets. This measurement uses data corresponding to an integrated luminosity of 9.7 fb$^{-1}$ collected by the \DO experiment. The measured integrated ratio is in agreement with predictions from NLO perturbative QCD and the Monte Carlo event generators {\sc alpgen} and {\sc pythia}. A good theoretical description of this process is essential since it forms a major background for a variety of physics processes, including standard model Higgs boson production in association with a $Z$ boson, $ZH(H\rightarrow b\bar{b})$, and searches for supersymmetric partners of the $b$ quark. These results will improve our theoretical understanding as we search for phenomena beyond the standard! model us! ing the data from similar collider experiments in the case where the final states of the interaction involved the production of vector bosons in association with two $b$-quark jets.\\ This work has been done in collaboration with \DO experiment but the analyses and results presented in this thesis are my contribution.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Advection of a passive vector field by the Gaussian velocity field with finite correlations in time

Using the field theoretic renormalization group technique the model of a passive vector field advected by an incompressible turbulent flow is investigated up to the second order of the perturbation theory (two-loop approximation). The turbulent environment is given by statistical fluctuations of the velocity field that has a Gaussian distribution with zero mean and defined noise with finite correlations in time. Two-loop analysis of all possible scaling regimes in general d-dimensional space is done in the plane of exponents ɛ − η, where ɛ characterizes the energy spectrum of the velocity field in the inertial range E ∝ k{sup 1−2ε}, and η is related to the correlation time at the wave number k which is scaled as k{sup −2+η}. It is shown that the scaling regimes of the present model of vector advection have essentially different properties than the scaling regimes of the corresponding model of passively advected scalar quantity. The results demonstrate the fact that within the present model of passively advected vector field the internal tensor structure of the advected field can have nontrivial impact on the diffusion processes deep inside in the inertial interval of given turbulent flow.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Model-Independent Searches for New Quarks at the LHC

New vector-like quarks can have sizable couplings to first generation quarks without conflicting with current experimental constraints. The coupling with valence quarks and unique kinematics make single production the optimal discovery process. We perform a model-independent analysis of the discovery reach at the Large Hadron Collider for new vector-like quarks considering single production and subsequent decays via electroweak interactions. An early LHC run with 7 TeV center of mass energy and 1 fb-1 of integrated luminosity can probe heavy quark masses up to 1 TeV and can be competitive with the Tevatron reach of 10 fb-1. The LHC with 14 TeV center of mass energy and 100 fb-1 of integrated luminosity can probe heavy quark masses up to 3.7 TeV for order one couplings.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Efficient gradient estimation using finite differencing and likelihood ratios for kinetic Monte Carlo simulations

While many optimization and control methods for stochastic processes require gradient information from the process of interest, obtaining gradient information from experiments is prohibitively expensive and time-consuming. As a result, such information is often obtained from stochastic process simulations. Computing gradients efficiently and accurately from stochastic simulations is challenging, especially for simulations involving computationally expensive models with significant inherent noise. In this work, we analyze and characterize the applicability of two gradient estimation methods for kinetic Monte Carlo simulations: finite differencing and likelihood ratio. We developed a systematic method for choosing an optimal perturbation size for finite differencing and discuss, for both methods, important implementation issues such as scaling with respect to the number of elements in the gradient vector. Through a series of numerical experiments, the methods were compared across different time and size regimes to characterize the precision and accuracy associated with each method. We determined that the likelihood ratio method is appropriate for estimating gradients at short (transient) times or for systems with small population sizes, whereas finite differencing is better-suited for gradient estimation at long times (steady state) or for systems with large population sizes.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Color Fluctuations in the Nucleon in High-Energy Scattering

We study quantum fluctuations of the nucleon's parton densities by combining QCD factorization for hard processes with the notion of cross section fluctuations in soft diffraction. The fluctuations of the small-x gluon density are related to the ratio of inelastic and elastic vector meson production in ep scattering. A simple dynamical model explains the HERA data and predicts the x and Q{sup 2} dependence of the ratio. In pp/pp scattering, fluctuations enhance multiple hard processes (but cannot explain the Tevatron CDF data), and reduce gap survival in central exclusive diffraction.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Inclusive τ lepton hadronic decay in vector and axial-vector channels within dispersive approach to QCD

The dispersive approach to QCD, which properly embodies the intrinsically nonperturbative constraints originating in the kinematic restrictions on relevant physical processes and extends the applicability range of perturbation theory towards the infrared domain, is briefly overviewed. The study of OPAL (update 2012) and ALEPH (update 2014) experimental data on inclusive τ lepton hadronic decay in vector and axial-vector channels within dispersive approach is presented.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Electrical design of the flexible imaging diffraction diagnostic for laser experiments (FIDDLE) at the National Ignition Facility (NIF)—Requirements, design, and performance

The Flexible Imaging Diffraction Diagnostic for Laser Experiments (FIDDLE) is a newly developed diagnostic for imaging time resolved diffraction in experiments at the National Ignition Facility (NIF). It builds on the successes of its predecessor, the Gated Diffraction Development Diagnostic (G3D). The FIDDLE was designed to support eight Daedalus version 2 sensors (six more hCMOS sensors than any other hCMOS-based diagnostic in NIF to date) and an integrated streak camera. We will review the electrical requirements, design, and performance of the electrical subsystems that were created to support this large number of cameras in the FIDDLE. The analysis of the data that the FIDDLE is intended to collect relies heavily on the accurate and well-understood timing of each sensor. We report camera-to-camera timing jitter of less than 100 ps rms and sensor integration times of 2.2 ns FWHM in 2-2 timing mode. Additionally, diffraction experiments on the NIF produce electric fields (EMI) on the order of 1 kV/m, which have been observed to negatively impact the performance of some electrical components of the FIDDLE. We report on the results of testing hCMOS camera electronics in a similar EMI environment generated in an offline lab. In addition, we also summarize the use of a novel approach to using a vector network analyzer as an EMI leak detector to understand and reduce the negative impacts of EMI on the FIDDLE.

47 OTHER INSTRUMENTATION↗

Algorithmic Improvements for Portable Event-Based Monte Carlo Transport Using the Nvidia Thrust Library

High performance computing environments are progressively moving towards many-core computing architectures. The Los Alamos National Laboratory Trinity machine, available in late 2016, will use both Intel Xeon Haswell processors and Intel Xeon Phi Knights Landing many integrated core (MIC) coprocessors. The Lawrence Livermore National Laboratory Sierra machine, available in 2018, will use an IBM PowerPC architecture along with Nvidia graphics processing units (GPUs). Applications that must work in this supercomputing environment must continue to adapt in order to take advantage of the diverse hardware architectures that are coming. A significant consideration is not only the performance of the application on a given platform but also the portability of the application to other platforms. The algorithmic improvements presented in this paper build upon recently-reported work on event-based Monte Carlo transport in the ALPSMC code that models particle transport in one-dimensional binary stochastic media. That paper discussed the lack of available vectorization in the traditional history-based algorithm used for Monte Carlo transport and presented a data parallel event-based algorithm implemented using the Nvidia Thrust library for portability. The performance of the data parallel event-based algorithm implemented using Thrust was compared to a native CUDA implementation. The conclusions from that work were that the Thrust library abstraction technique caused too significant a loss in performance but that the event-based method was a viable option that should be further investigated. In this paper, we describe algorithmic improvements to the data parallel event-based algorithm previously presented. We made further algorithmic optimizations to the event-based CUDA implementation, most notably: data structure changes, a new conditional particle removal scheme in the event-based process, and the use of multiple GPUs. In addition to improvements to the algorithm, we re-implemented the Thrust version from the now further optimized CUDA version, giving a greater chance for success at a performant abstraction. Finally, we revisited our previous assumptions about the inability of the history-based method to achieve performance on vector style architectures such as the MICs and GPUs, with surprising and promising results. (authors)

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Connections between nonlocal operators: From vector calculus identities to a fractional Helmholtz decomposition

Nonlocal vector calculus, which is based on the nonlocal forms of gradient, divergence, and Laplace operators in multiple dimensions, has shown promising applications in fields such as hydrology, mechanics, and image processing. In this work, we study the analytical underpinnings of these operators. We rigorously treat compositions of nonlocal operators, prove nonlocal vector calculus identities, and connect weighted and unweighted variational frameworks. We combine these results to obtain a weighted fractional Helmholtz decomposition which is valid for sufficiently smooth vector fields. Our approach identifies the function spaces in which the stated identities and decompositions hold, providing a rigorous foundation to the nonlocal vector calculus identities that can serve as tools for nonlocal modeling in higher dimensions.

97 MATHEMATICS AND COMPUTING↗

Graphic contrastive learning analyses of discontinuous molecular dynamics simulations: Study of protein folding upon adsorption

A comprehensive understanding of the interfacial behaviors of biomolecules holds great significance in the development of biomaterials and biosensing technologies. In this work, we used discontinuous molecular dynamics (DMD) simulations and graphic contrastive learning analysis to study the adsorption of ubiquitin protein on a graphene surface. Our high-throughput DMD simulations can explore the whole protein adsorption process including the protein structural evolution with sufficient accuracy. Contrastive learning was employed to train a protein contact map feature extractor aiming at generating contact map feature vectors. Subsequently, these features were grouped using the k-means clustering algorithm to identify the protein structural transition stages throughout the adsorption process. The machine learning analysis can illustrate the dynamics of protein structural changes, including the pathway and the rate-limiting step. Our study indicated that the protein–graphene surface hydrophobic interactions and the π–π stacking were crucial to the seven-stage adsorption process. Upon adsorption, the secondary structure and tertiary structure of ubiquitin disintegrated. The unfolding stages obtained by contrastive learning-based algorithm were not only consistent with the detailed analyses of protein structures but also provided more hidden information about the transition states and pathway of protein adsorption process and structural dynamics. Our combination of efficient DMD simulations and machine learning analysis could be a valuable approach to studying the interfacial behaviors of biomolecules.

97 MATHEMATICS AND COMPUTING↗

Resonance energy transfer: The unified theory via vector spherical harmonics

In this work, we derive the well-established expression for the quantum amplitude associated with the resonance energy transfer (RET) process between a pair of molecules that are beyond wavefunction overlap. The novelty of this work is that the field of the mediating photon is described in terms of a spherical wave rather than a plane wave. The angular components of the field are constructed in terms of vector spherical harmonics while Hankel functions are used to define the radial component. This approach alleviates the problem of having to select physically correct solution from non-physical solutions, which seems to be inherent in plane wave derivations. The spherical coordinate system allows one to easily decompose the photon’s fields into longitudinal and transverse components and offers a natural way to analyse near-, intermediate-, and far-zone RET within the context of the relative orientation of the transition dipole moments for the two molecules.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Monitoring by Use of Clusters of Sensor-Data Vectors

The inductive monitoring system (IMS) is a system of computer hardware and software for automated monitoring of the performance, operational condition, physical integrity, and other aspects of the health of a complex engineering system (e.g., an industrial process line or a spacecraft). The input to the IMS consists of streams of digitized readings from sensors in the monitored system. The IMS determines the type and amount of any deviation of the monitored system from a nominal or normal ( healthy ) condition on the basis of a comparison between (1) vectors constructed from the incoming sensor data and (2) corresponding vectors in a database of nominal or normal behavior. The term inductive reflects the use of a process reminiscent of traditional mathematical induction to learn about normal operation and build the nominal-condition database. The IMS offers two major advantages over prior computational monitoring systems: The computational burden of the IMS is significantly smaller, and there is no need for abnormal-condition sensor data for training the IMS to recognize abnormal conditions. The figure schematically depicts the relationships among the computational processes effected by the IMS. Training sensor data are gathered during normal operation of the monitored system, detailed computational simulation of operation of the monitored system, or both. The training data are formed into vectors that are used to generate the database. The vectors in the database are clustered into regions that represent normal or nominal operation. Once the database has been generated, the IMS compares the vectors of incoming sensor data with vectors representative of the clusters. The monitored system is deemed to be operating normally or abnormally, depending on whether the vector of incoming sensor data is or is not, respectively, sufficiently close to one of the clusters. For this purpose, a distance between two vectors is calculated by a suitable metric (e.g., Euclidean distance) and "sufficiently close" signifies lying at a distance less than a specified threshold value. It must be emphasized that although the IMS is intended to detect off-nominal or abnormal performance or health, it is not necessarily capable of performing a thorough or detailed diagnosis. Limited diagnostic information may be available under some circumstances. For example, the distance of a vector of incoming sensor data from the nearest cluster could serve as an indication of the severity of a malfunction. The identity of the nearest cluster may be a clue as to the identity of the malfunctioning component or subsystem. It is possible to decrease the IMS computation time by use of a combination of cluster-indexing and -retrieval methods. For example, in one method, the distances between each cluster and two or more reference vectors can be used for the purpose of indexing and retrieval. The clusters are sorted into a list according to these distance values, typically in ascending order of distance. When a set of input data arrives and is to be tested, the data are first arranged as an ordered set (that is, a vector). The distances from the input vector to the reference points are computed. The search of clusters from the list can then be limited to those clusters lying within a certain distance range from the input vector; the computation time is reduced by not searching the clusters at a greater distance.

Iverson, David L.↗

Status and Examples for the Version 06 IMERG Multi-Satellite Products

After five years of development following the launch of the Global Precipitation Measurement (GPM) missionCore Observatory, the GPM data products are now being extended across the joint Tropical Rainfall MeasuringMission (TRMM) and GPM eras. Version 06 of the U.S. GPM team's Integrated Multi-satellitE Retrievals forGPM (IMERG) merged precipitation product provides a consistent intercalibration for all precipitation productscomputed from individual satellites with the TRMM and GPM Core Observatory sensors as the TRMM- andGPM-era calibrators, respectively, and incorporates monthly surface gauge data. One major change in the basicIMERG algorithm for V06 is that precipitation motion vectors (used to drive the quasi-Lagrangian interpolation,or "morphing") are computed by tracking vertically integrated vapor (TQV) fields analyzed in MERRA2 andGEOS5. This innovation provides globally complete coverage, expanding IMERG's coverage beyond the 60N-Slatitude band previously provided by IR-based vectors, although precipitation over snowy/icy surfaces is stillmasked out as unreliable. A second innovation is that the Quality Index (QI) data field computed for the half-hourlydatasets has been refined to include estimates of correlation at microwave overpass times.We will summarize the processing status for V06 IMERG, for which the retrospective processing shouldbe actively advancing at meeting time. We will show early examples of performance. For example, the TQVmotion vectors are typically slightly better than the IR-based vectors at all latitudes. The transition across theTRMM/GPM data boundary will be discussed, including the necessity of filling in the TRMM-based calibrationsover the latitude band 35-65 in each hemisphere. The notional schedule for the eventual retirement of thepredecessor TRMM Multi-satellite Precipitation Analysis (TMPA) multi-satellite dataset will be updated as well.

Huffman, George J.↗

Kink mechanism in Cu/Nb nanolaminates explored by $\mathcal{in}$ $\mathcal{situ}$ pillar compression

We report Nano metallic laminates (NMLs) exhibit different failure modes depending on the loading conditions due to their mechanical anisotropies. Kinking is a typical failure mode in many NMLs compressed along a layer-parallel direction. However, a detailed description of the microstructure evolution during kink band (KB) formation and an in-depth understanding of the formation mechanisms are lacking. In this work, the KB process is investigated in Cu/Nb NMLs by in situ micro pillar compression in the scanning electron microscope (SEM) along a layer-parallel direction. Post-mortem S/TEM and transmission Kikuchi diffraction (TKD) analyses show that kink banding leads to significant microstructure changes characterized by an accumulation of geometrically necessary dislocations (GNDs) and of tilt geometrically necessary boundaries (GNBs) near KB boundaries (KBBs). The distinct microstructure evolution implies that KB formation is facilitated by the inhomogeneous microstructures resulting in constrained deformation modes. Specifically, dislocations active on slip planes nearly parallel to the interfaces make a major contribution to kink evolution after the onset of kinking. Once layer-parallel slip systems are activated, preexisting lattice dislocations and dislocations nucleating from interfaces will accumulate as GNDs near KBBs via the stochastic storage of lattice dislocations that have certain Burgers vectors. GNDs can further transform into GNBs via cross-slip and climb driven processes near the KBB. Furthermore, GNBs near KBBs can grow by incorporating more GNDs or by coalescence to accommodate the KB evolution. We further hypothesize that microstructural perturbations and their ensuing stresses can initiate KB formation in Cu/Nb NMLs.

36 MATERIALS SCIENCE↗

Sparse Binary Matrix-Vector Multiplication on Neuromorphic Computers

Neuromorphic computers offer the opportunity for low-power, efficient computation. Though they have been primarily applied to neural network tasks, there is also the opportunity to leverage the inherent characteristics of neuromorphic computers (low power, massive parallelism, collocated processing and memory) to perform non-neural network tasks. Here, we demonstrate how an approach for performing sparse binary matrix-vector multiplication on neuromorphic computers. We describe the approach, which relies on the connection between binary matrix-vector multiplication and breadth first search, and we introduce the algorithm for performing this calculation in a neuromorphic way. We validate the approach in simulation. Finally, we provide a discussion of the runtime of this algorithm and discuss where neuromorphic computers in the future may have a computational advantage when performing this computation.

Schuman, Catherine↗