Serial Computer Research Articles

The Influence of Parallel Computing on Building Deep Learning Model for the Classification of Bean Diseases

In recent years, the utilization of deep learning techniques for image classification has made significant strides in the field of agriculture. One of the key areas of interest in agriculture is the early detection and classification of diseases in crops, as this can have an insightful impact on crop revenue and quality. This research has investigated the influence of parallel computing on the performance of a deep learning-based classification model for diagnosing bean diseases. Specifically, we have explored the use of parallel computing frameworks to accelerate model training and inference, thereby enhancing the efficiency and effectiveness of disease classification. Our findings demonstrated the potential for parallel computing to accelerate model training. When training a bean disease classification model, we achieved an accuracy of 0.93 using parallel computing, compared to 0.83 with serial computing. Moreover, parallel computing significantly reduced training time, taking only 3 minutes compared to 51 minutes with serial computing.

Vectfem: a generalized MATLAB-based vectorized algorithm for the computation of global matrix/force for finite elements of any type and approximation order in linear elasticity

In this paper, we introduce a new vectorized MATLAB-based algorithm for efficient serial computation of global matrix/force arising from finite element method (FEM) for meshes of any type and approximation order in linear elasticity. Because for-loops in MATLAB are very slow, we propose a modified process that takes advantage of vectorization and sparse assembly to achieve good performance while using the same memory as the standard algorithm. For this purpose, by using good programming practices, the implementation of this scheme is succinctly described and can be integrated into any MATLAB package dealing with FEM. Specifically, attention is paid to the calculation of the triplet (row index, column index, matrix components) as well as the assembly of the global stiffness matrix, mass matrix and force vector. Additionally, an extension of the proposed approach for Mindlin plate theory and functionally graded materials is outlined. Finally, the accuracy of this strategy is verified on selected numerical tests after comparing the obtained results with those of ABAQUS. In terms of performance, the study conducted on a set of meshes considering the standard algorithm and two other well-known MATLAB vectorized algorithms revealed that: (i) for a 2D beam problem meshed with P1\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$P_{1}$$\\end{document}-triangle elements, a speedup of about 8 and 15 is achieved with sparse and fsparse, respectively. (ii) for a 3D plate problem meshed with P1\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$P_{1}$$\\end{document}-tetrahedral elements, a speedup of about 4 and 8 is achieved with sparse and fsparse, respectively. When compared to ABAQUS performance, the proposed scheme results in a computational time that is about five times smaller.

ВКЛАД ОТЕЧЕСТВЕННЫХ УЧЕНЫХ В РАЗВИТИЕ УСТРОЙСТВ ВЫЧИСЛИТЕЛЬНОЙ ТЕХНИКИ

The study reflected two stages of digital computing, when discrete counting devices and analog computers were created. (Research purpose) The research purpose is identifying the design features of computers in which the ideas of domestic scientists were implemented at the mechanical and electromechanical stages of the development of computer technology. (Materials and methods) Conducted a historical and scientific analysis of literary sources, original works of Russian scientists, descriptions of inventions and patents related to the mechanical and electromechanical stage of the development of computer technology. We studied materials from the Archive of the USSR Academy of Sciences, descriptions of serial computers manufactured in the USSR. (Results and discussion) We considered the outstanding role played by academician P.L. Chebyshev in the history of the construction of continuous counting machines. The design features of the adding machine created by V.T. Odner and further developments of adding machines by domestic engineers (adding machine "Soyuz") were presented. It was noted that with the invention of the Odner adding machine in 1890, the stage of serial production of computer equipment in Russia began. It was shown that in the field of analog computing, a significant contribution was made in the construction of hydraulic integrators by engineer V.S. Lukyanov; electrical integrators by academicians A.N. Krylov, I.S. Bruk. The first electronic analog computers were created at the Electrotechnical Institute of the USSR Academy of Sciences under the leadership of L.I. Gutenmacher. (Conclusions) The ideas and design features of mechanical and analog computing devices were reflected in the works of domestic scientists and engineers. We have traced the continuity in the creation of mechanical and electromechanical calculating machines. We have studied unique samples of computer technology, such as the integrator by A.N. Krylov, the hydraulic integrator by V.S. Lukyanov, the electrical integrators by I.S. Bruk, L.I. Gutenmacher, the electro-optical reading machine by engineer V.E. Agapov.

An agent‐based model of consumer demand

AbstractThis paper studies an agent‐based model of consumer demand. Agents are heterogeneous with respect to their preferences and incomes. There are two basic ingredients in the model. The first ingredient is a metric that captures the degree of heterogeneity between agents. The second ingredient is a serial computer algorithm that is used in order to compute a terminal consumption bundle at which income is exhausted and overall utility is maximized. Agents are clustered into heterogeneous groups based on their preferences and incomes. We extract information about the evolution of consumer expenditure under different price regimes and the buildup of optimal demand for varying levels of income and preference parameter values. These features cannot be obtained in the classical framework of static utility maximization. Our agent‐based data‐driven methodology can be applied to any relevant data set and so provide a reliable model for forecasting demand given some agent characteristics.

A Multi-Time-Step Parallel Computing Method based on Overlapping Particles for DEM

The application of the discrete element method (DEM) to continuous medium problems is becoming increasingly widespread. In this work, a parallel computing method with multiple time steps based on overlapping particles is proposed. The domain decomposition method (DDM) with overlapping particles method is used to increase the speed up and shorten the computation time and meet the consistency requirements during data transmission. The multi-time-step method (MTSM) is adopted to tackle the matching of asynchronous step boundary. Data are exchanged at the boundaries in each subdomain with a message passing interface (MPI). The computational efficiency of different step ratios in both serial and parallel computing is studied respectively. Numerical examples show that the DEM can effectively handle large structure deformation problems, and provides a shorter calculation time than that of the finite element method (FEM). The DEM with multiple time steps in different subdomains effectively reduces the computation time than that with a single time step in the entire domain. Under fixed step ratio conditions, using parallel computing can save more time than serial computing. This work develops ideas for expanding the application of DEM for large engineering problems.

Optimal dispatch for cross-regional integrated energy system with renewable energy uncertainties: A unified spatial-temporal cooperative framework

Cross-regional energy sharing is a promising way to alleviate the imbalance in energy distribution. However, there is a risk that the uncertain fluctuations of intermittent renewable energy sources (RES) could be transmitted across regional transmissions, resulting in large-scale system economic inefficiency and security concerns. To this end, a unified spatial-temporal cooperative framework for the integrated energy system, which considers the coordination between intra-regional multi-energy coupling and cross-regional multi-agent-system energy sharing, is proposed in this paper. First, in the temporal dimension, a multi-time-scale stochastic decision-making model is established based on stochastic model predictive control (SMPC). Thus, slow-response units are arranged to gain economic goals in day-ahead stage and the closed-loop rolling adjustment is implemented on fast-response units to achieve accuracy supply-demand balance in real-time stage. Second, in the spatial dimension, a fully distributed dispatch approach is adopted to address the information sharing barriers, which only requires neighboring information and considers consistency of boundary control variables on border lines. Finally, to improve the computational efficiency, the established complex time sequential model is decomposed into a set of time-discrete parallel sub-problems, using the optimal condition decomposition (OCD) technique. Simulation results on a cross-regional test system demonstrate that proposed framework can increase the wind power consumption capacity and reduce operation cost through cross-regional cooperation, while the parallel decomposition can save 67.25 % of the computation time relative to the serial computation.

A new active learning method for reliability analysis based on local optimization and adaptive parallelization strategy

In recent years, an active learning method combining Kriging and Monte Carlo Simulation (AK-MCS), has been developed for calculating the failure probability. However, the original AK-MCS only uses serial computing, which limits its ability to take advantage of distributed computing. Thus, this work introduces a novel adaptive learning approach for reliability analysis by combining local optimization and a parallelization strategy. The new approach identifies the points of greatest uncertainty through local optimization and adds them into the design of experiments. An inner learning loop is implemented, searching for additional best points with a pseudo Kriging model so that the performance function can be evaluated in parallel. Furthermore, an adaptive strategy is proposed to determine the amount of additional points during iteration based on the minimum value of the learning function. We conducted a comparison between the proposed method and the original AK-MCS as well as a few additional methods in order to assess its efficacy and precision. Five examples were considered to assess performance.

ALLIANCE: Spectral solver for kinetic plasma turbulence

The ALLIANCE (ALLIANCE - spectrAL soLver for kInetic plAsma turbuleNCE) code is developed to solve a new set of four-dimensional electromagnetic drift-kinetic equations in slab geometry [1]. The nonlinear equations are useful for the study of magnetized plasma systems at scales comparable to, or larger than the ion gyroradius. In particular, it is suited for the study the kinetic turbulent cascade in astrophysical plasma, while preserving finite Larmor radius effects at the fluid-kinetic transition. The equations solved are in spectral Fourier-Laguerre-Hermite form, a pseudo-spectral approach is used for the nonlinear terms, and the code is parallelised over multiple directions. After a presentation of the code, validation runs are shown, and benchmarks for serial and parallel computations are presented.

PGRASS-Solver: A Graph Spectral Sparsification-Based Parallel Iterative Solver for Large-Scale Power Grid Analysis

With the increase in the complexity of VLSI chips, power grid analysis has become a challenging task, because linear equations of extremely large size need to be solved. Recent graph sparsification based solvers have shown promising performance for power grid analysis. However, existing graph sparsification algorithms are implemented in serial computing, while factorization and backward/forward substitution of the sparsifier’s Laplacian matrix are hard to parallelize. On the other hand, partition based iterative methods which are inherently parallel lack a direct control of the relative condition number of the preconditioner and consume more memory. In this work, we propose a novel parallel iterative solver called pGRASS-Solver. We first propose a practically-efficient parallel graph sparsification algorithm. Then domain decomposition method (DDM) is utilized to solve the sparsifier’s Laplacian matrix. To further improve the efficiency, a variant of DDM which employs partial Cholesky factorization and Schur complement matrix sparsification is proposed. Thus, we obtain an efficient parallel preconditioner, which not only leads to fast convergence but also enjoys ease of parallelization. Numerous experiments are conducted to illustrate the superior efficiency of the proposed pGRASS-Solver for large-scale power grid analysis, showing an average 6.8X speedup over a recent parallel iterative solver [1]. Moreover, it solves a real-world power grid matrix with 0.36 billion nodes and 8.7 billion nonzeros within 20 minutes on a 16-core machine, which is 10.9X faster than the best result of sequential graph sparsification based solver [2].

Accelerate the Time-Reversal Computing of Lightning Location Using GPU and Phase-Difference Filter

Time-reversal (TR) technique has been proved to be an effective method to locate the lightning sources, which could get much more accurate results and is capable of finding the weak radiation sources. However, the huge amount of calculation makes it not meet the fast localization requirement of the lightning map array. To accelerate the localization speed of TR, this article proposes to use graphics processing unit (GPU) parallel computing and shared memory to calculate the radiation source directions. Then, the phase difference is combined with TR to filter out the less relevant direction points thus to further speed up the computation. Finally, as for the real lightning data received by the antennas, not every sliding window contains the radiation signals during the TR calculation, we use the signal-interference energy ratio to judge whether a sliding window has the radiation signal, thus, we only need to calculate the windows of which the ratios are above the threshold. Experiments show that, in simulation, by using the first two skills, the computing speed is greatly improved by more than 1700 times than a single CPU serial computing; in 500 ms real lightning data localization, our method is 263 times faster than the original CPU methods, meanwhile, the location result is as good as the latter.

Improved fast iterative method for higher calculation accuracy of traveltimes

Fast Iterative Method (FIM), a grid-based ray-tracing algorithm, has a computational efficiency far exceeding that of conventional fast marching method (FMM) in serial computing, because it adopts an iterative algorithm and can be calculated in parallel. However, the finite difference scheme of the FIM only employs the first-order Godunov upwind finite difference scheme, resulting in low calculation accuracy of arrival traveltimes. This paper proposes a new algorithm of FIM, which utilizes second-order and mixed-order difference schemes under the parallel computing structure, effectively improves the calculation accuracy of traveltimes, and basically maintains the original computational efficiency. Especially, improved finite difference calculation in the diagonal direction of square grids and the double-grid technique near the source are applied to further reduce overall calculation error. Numerical experiments show that for the given model, the calculation error can be reduced from 2 ms of the original FIM calculation to 0.05 ms, and the calculation accuracy is increased by 40 times, while the time consumption increases only 27%.

Fundamental energy cost of finite-time parallelizable computing

The fundamental energy cost of irreversible computing is given by the Landauer bound of kTln 2/bit, where k is the Boltzmann constant and T is the temperature in Kelvin. However, this limit is only achievable for infinite-time processes. We here determine the fundamental energy cost of finite-time parallelizable computing within the framework of nonequilibrium thermodynamics. We apply these results to quantify the energetic advantage of parallel computing over serial computing. We find that the energy cost per operation of a parallel computer can be kept close to the Landauer limit even for large problem sizes, whereas that of a serial computer fundamentally diverges. We analyze, in particular, the effects of different degrees of parallelization and amounts of overhead, as well as the influence of non-ideal electronic hardware. We further discuss their implications in the context of current technology. Our findings provide a physical basis for the design of energy-efficient computers.

A Parallel Principal Skewness Analysis and Its Application in Radar Target Detection

Radar is often affected by various clutter backgrounds in complex environments, so clutter suppression has important practical significance for radar target detection. The clutter suppression process conforms to the blind source separation (BSS) model. The principal skewness analysis (PSA) algorithm is a BSS algorithm with third-order statistics as the objective function, and its running speed is faster than the conventional BSS algorithm. Still, the PSA algorithm has the problem of error accumulation. This paper improves the PSA algorithm and proposes a parallel PSA (PPSA) algorithm. PPSA can estimate the directions corresponding to each independent component simultaneously and avoid the problem of error accumulation. PPSA uses parallel instead of serial computing, significantly improving the running speed. In this paper, the PPSA algorithm is applied to radar target detection. The simulation data and real data experiments verify the effectiveness and superiority of the PPSA algorithm in suppressing clutter.

Serial and Parallel Computation of Bone Scan Image Processing

Serial and Parallel Computation of Bone Scan Image Processing

8-b Precision 8-Mb ReRAM Compute-in-Memory Macro Using Direct-Current-Free Time-Domain Readout Scheme for AI Edge Devices

Compute-in-memory (nvCIM) macros based on non-volatile memory make it possible for artificial intelligence (AI) edge devices to perform energy-efficient multiply-and-accumulate (MAC) operations by minimizing the movement of data between the processors and memory. However, nvCIM imposes tradeoffs between energy efficiency, computing latency, and readout accuracy against process variation. To overcome these challenges, this work proposed a nvCIM macro featuring: 1) a direct-current-free time-space-based in-memory computing (DCFTS-IMC) scheme; 2) a wordline-based serial access computing (WSAC) scheme; 3) an integration-based voltage-to-time converter (IVTC); and 4) a hidden-latency time-to-MAC value conversion (HLTMC) scheme. The proposed 22-nm 8-Mb resistive random access memory-CIM (ReRAM-CIM) macro was fabricated to demonstrate MAC operations with 8-b input, 8-b weight, and 19-b output. Our nvCIM macro achieved computing latency of 14.4 ns under 8-b precision with an energy efficiency of 21.6 TOPS/W.

Dynamic complex opto-magnetic holography

Despite recent significant progress in real-time, large-area computer-generated holography, its memory requirements and computational loads will be hard to tackle for several decades to come with the current paradigm based on a priori calculations and bit-plane writing to a spatial light modulator. Here we experimentally demonstrate a holistic approach to serial computation and repeatable writing of computer-generated dynamic holograms without Fourier transform, using minimal amounts of computer memory. We use the ultrafast opto-magnetic recording of holographic patterns in a ferrimagnetic film with femtosecond laser pulses, driven by the on-the-fly hardware computation of a single holographic point. The intensity-threshold nature of the magnetic medium allows sub-diffraction-limited, point-by-point toggling of arbitrarily localized magnetic spots on the sample, according to the proposed circular detour-phase encoding, providing complex modulation and symmetrical suppression of upper diffractive orders and conjugated terms in holographically reconstructed 3-D images.

A fast identification method of train axle crack parameters based on semi-inverse Kriging method

On-line monitoring of crack parameters can prevent axle breaking accidents and potentially disastrous consequences, while satisfying the requirement of the damage tolerance, and improving the economy in operation. A fast and reliable crack parameters identification method is required for on-line monitoring of a large number of axles. Based on the response surface characteristics of crack position and depth at two sensors, a new method is proposed adopting two Kriging surrogate models constructed by the 2xRev values of vibration signals from two measurement points. Firstly, by exchanging part of inputs and outputs, a semi-inverse Kriging model is constructed to obtain a crack equivalent line. Then, the crack equivalent line is projected on another Kriging model as the convergence path, which is the key step transforming a two-objective optimization problem into a one-dimensional searching problem. Finally, an improved dichotomy method is used to obtain the result, by which the total number of steps and the calculation time is controlled on the premise of ensuring accuracy. Different from current multi-objective optimization algorithms searching with parallel computing, the proposed method reduces the number of parameters and searching model space to be one, belonging to serial computing. The comparison to other methods shows excellent features in computing speed and stability, and the experiment was implemented on a double-disk rotor system with similar structure to a wheelset, validating the proposed methodology.

Solving the 3‐Satisfiability Problem Using Network‐Based Biocomputation

The 3‐satisfiability Problem (3‐SAT) is a demanding combinatorial problem that is of central importance among the nondeterministic polynomial (NP) complete problems, with applications in circuit design, artificial intelligence, and logistics. Even with optimized algorithms, the solution space that needs to be explored grows exponentially with the increasing size of 3‐SAT instances. Thus, large 3‐SAT instances require excessive amounts of energy to solve with serial electronic computers. Network‐based biocomputation (NBC) is a parallel computation approach with drastically reduced energy consumption. NBC uses biomolecular motors to propel cytoskeletal filaments through nanofabricated networks that encode mathematical problems. By stochastically exploring possible paths through the networks, the cytoskeletal filaments find possible solutions. However, to date, no NBC algorithm for 3‐SAT has been available. Herein, an algorithm that converts 3‐SAT into an NBC‐compatible network format is reported and four small 3‐SAT instances (with up to three variables and five clauses) using the actin–myosin biomolecular motor system are experimentally solved. Because practical polynomial conversions to 3‐SAT exist for many important NP complete problems, the result opens the door to enable NBC to solve small instances of a wide range of problems.

The collocation spectral method with domain decomposition for radiative heat transfer in two-dimensional enclosures

In this paper, the collocation spectral method (CSM) combined with domain decomposition method is developed to solve the radiative transfer equation in two-dimensional irregular domains. Three benchmark problems consist of the square enclosure, the L-shaped enclosure and the square enclosure with a centered obstruction are solved and compared with the published data to validate the ability of present developed method. The comparison shows good agreements and indicates that the method has a good accuracy for all problems. Then, the performances of influence matrix technique and iterative substructuring technique to exchange the radiative information between subdomains are compared. It is found that, in the serial computations, the computational cost of influence matrix technique is hundreds of times more expensive than that of iterative substructuring technique. The high cost of influence matrix technique is due to that the radiative intensity is a high dimensional variable with angular dependence, and tremendous subproblems have to be solved to construct the influence matrix. Finally, the modified CSM with domain decomposition, in which the radiative intensity is decomposed into three components, the real wall-related one, the virtual shared interface-related one, and the medium-related one, is proposed. The first two components are solved analytically, and the last one is still solved by the CSM. With such treatments, the computational cost is slightly increased, but the ray effect originated from step-change temperature of medium or step-change optical parameters can be effectively mitigated. Besides, the modified CSM with domain decomposition can also avoid the ray effect due to shadowing singularities. But it should be noted that, in such case, the ray effect due to inhomogeneous spatial distribution of source term may emerge. In conclusion, the CSM combined with domain decomposition is a good alternative method for thermal radiation calculation in complex geometry which can be decomposed into regular subdomains.

A fast serial computation approach for computing viewshed of a region by detecting key spots on the terrain

Computing a viewshed for a space (a polygon), rather than a single location (a point), requires analysis of the maximum terrain extent observable from any location within the space. For a digital elevation model (DEM), it is the sum of viewshed analyses for all the pixels inside the confined space. As the raster spatial resolution increases, the process becomes computationally expensive. In this study, we propose a new serial computation approach, named Regional Viewshed by Aspect Peaks (RVAP), in which key pixels that define the cumulative visibility map on a digital elevation model (DEM) are found, making the process computationally affordable. We compare our proposed method next to 1) a 5-cell grid interval method 2) and an all-encompassing but slow complete approach where the viewshed is computed for all the pixels and then overlaid. Six regions-of-interests (ROIs) on two Shuttle radar (∼555 × 555 km dimensions, 90-meter resolution) DEMs are selected and categorized by their terrain ruggedness index (TRI) (i.e. slightly rugged, mildly rugged, highly rugged). The proposed approach is much faster, while its coverage is identical to the all-encompassing approach. Moreover, as the TRI increases, the accuracy of the proposed algorithm is not diminished, while its execution time decreases.