Acceleration Algorithm Research Articles

An enhanced branch-and-bound algorithm for bilevel integer linear programming

Bilevel integer linear programming (BILP) problems have been studied for decades. Many exact algorithms have been proposed in recent years for small- or medium-sized instances. However, few of these algorithms were shown to be efficient on large-sized instances. In this paper, we present an enhanced branch-and-bound algorithm for a class of BILP problems, which can discard a subspace from the search space in each iteration larger than that in a benchmark branch-and-bound algorithm. The corresponding enhanced branching rule can efficiently slow down the creation of new node problems so as to significantly reduce the computation time. Our scheme may be suboptimal if the lower-level problem is not unique optimal as the enhanced branching rule may discard bilevel feasible solutions that may turn out to be optimal to the bilevel programming. We present computational studies to evaluate the algorithm speedup and solution quality of our algorithm, compared with state-of-the-art algorithms from the literature on a large testbed of general BILP instances, some of which are still unsolved. The computational results show that our enhanced branching rule can achieve significant speedup on the benchmark branching rule with satisfying solution quality. In particular, our algorithm shows superior performance on large-sized BILP instances with a relatively complex lower-level problem.

An Edge Storage Acceleration Service for Collaborative Mobile Devices

Fueled by the advances in the Internet of Things, and the growing capacity of smart mobile devices at the edge of the Internet, we have witnessed a growing trend in research and development for edge computing and edge storage, which extends the abilities of single mobile device on the edge through on-demand collaboration among multiple geographically distributed mobile devices. In this article, we address several technical challenges that are unique for collaborative storage at the edge due to the unique characteristics of mobile devices. First, we formalize the collaborative storage problem as an optimization problem. Second, we design an Acceleration Algorithm for Collaborative Storage, called A2CS, based on the architecture of Alternating Direction Method of Multipliers (ADMM). Specifically, we use the Nesterov’s Acceleration strategy and the step size rules in the process of updating variables and determining the optimal speed of convergence. We develop a novel collaborative storage policy in order to guide the whole lifecycle of collaborative storage. Finally, we conduct a series of experiments for acceleration performance analysis and validation. We show that A2CS delivers a better convergence performance with different step size rules, compared with two existing approaches: the ADMM baseline and the ADMM-OR (ADMM with Over-Relaxation), achieving the acceleration percentage by at least 25.33 percent and at most 64.01 percent. In addition, by conducting the utility performance comparison analysis with the existing Average Distribution Strategy (ADS) and the existing Distance Preferred Distribution Strategy (DPDS), we show the advantage of A2CS over both ADS and DPDS with respect to the total utility and energy consumption.

Minimum Cut Acceleration by Exploiting Tree-Cut Injection for Upper Bound Estimation

As a basic problem in graph theory, the minimum cut (min-cut) problem has important applications in communication related researches. There are injections from the set of all traversal trees to the set of all cuts. However, this kind of injection is completely ignored in existing methods, which may lead to the lose of acceleration opportunity. Accordingly, we propose Traversal Tree-Cut Injection based Min-Cut Acceleration algorithm (TCMA). By collecting enough cuts from different traversal trees in an undirected graph, the upper bound of min-cut value between any node pair can be calculated as the minimal value of all related cuts, which can match the accurate min-cut value with high probability. We compare TCMA with other methods using random and real-world graphs. The results show that different from other methods, TCMA is general enough to not rely on any special topology, and can achieve both high acceleration ratio and precision in various graphs, thus can be used as an effective supplement to existing methods.

Design and Function Realization of Nuclear Power Inspection Robot System

SUMMARYThe particularity of nuclear power plant environment requires that the nuclear power inspection robot must be remote control operation. The main purpose of the inspection robot is to carry out inspection, prevention, reporting, and safety emergency operation on the instruments, so as to provide guarantee for the safe operation of the nuclear power plant. Based on the representative configuration of nuclear power robot at home and abroad, this paper develops a small and lightweight nuclear power plant inspection robot, including walking mechanism, lifting mechanism, operating mechanism, image acquisition, information communication and control system, etc., to carry on the statics analysis to the key components of the inspection robot and verify that the stiffness and strength of the mechanical structure meet the requirements of lightweight design. Modal analysis is carried out to verify that the motor does not cause resonance when working. The kinematic model of the robot has been established and can provide the theoretical basis for the controller design. A hierarchical control system based on LabVIEW upper computer monitoring and control operation interface is established, which uses adaptive fuzzy Proportional Integral Derivative (PID) control to simulate the walking control, and then realizes the control of walking mechanism through software programming, and the adaptive fuzzy PID control has better effect than the conventional PID control. The S-type acceleration and deceleration algorithm is used to realize the accurate control of the position location of the lifting mechanism. Finally, combined with the experiment of 5MS robot comprehensive experimental platform, it is proved that the inspection robot can realize remote control function operation.

On the convergence of exact distributed generalisation and acceleration algorithm for convex optimisation

In this paper, we study distributed multiagent optimisation over undirected graphs. The optimisation problem is to minimise a global objective function, which is composed of the sum of a set of local objective functions. Recent researches on this problem have made significant progress by using primal-dual methods. However, the inner link among different algorithms is unclear. This paper shows that some state-of-the-art algorithms differ in that they incorporate the slightly different last dual gradient terms based on the augmented Lagrangian analysis. Then, we propose a distributed Nesterov accelerated optimisation algorithm, where a doubly stochastic matrix is allowed to use, and nonidentical local step-sizes are employed. We analyse the convergence of the proposed algorithm by using the generalised small gain theorem under the assumption that each local objective function is strongly convex and has Lipschitz continuous gradient. We prove that the sequence generated by the proposed algorithm linearly converge to an optimal solution if the largest step-size is positive and less than an explicitly estimated upper bound, and the largest momentum parameter is nonnegative and less than an upper bound determined by the largest step-size. Simulation results further illustrate the efficacy of the proposed algorithm.

A finite volume method preserving maximum principle for steady heat conduction equations with modified Anderson acceleration

In this article, we propose a finite volume scheme preserving the discrete maximum principle (DMP) for steady heat conduction equations on distorted meshes. In contrary to these finite volume schemes preserving DMP, our new scheme uses the geometric average (instead of harmonic average) of two one-side numerical heat fluxes, especially it produces a more accurate flux approximation, which is verified numerically. We prove that there hold the DMP and the existence of a solution for our scheme. We also propose a modified Anderson acceleration (MAA) algorithm to improve the robustness and accelerate the convergence. The algorithm design is based on a minimization problem for a linear combination of the residual vectors of the nonlinear system. Numerical experiments verify the DMP-preserving property of our scheme and the efficiency of the MAA algorithm. Moreover, the stability and efficiency of the Picard iteration with MAA are much better than that with the classical Anderson acceleration. In the numerical examples, the convergence rate of MAA iteration is up to seven times of the convergence rate of the Picard iteration.

Modeling Superconductor SFN-Structures Using the Finite Element Method

We consider the problem of mathematical modeling of the current distribution in Josephson structures based on semiclassical equations of the microscopic theory of superconductivity (the Usadel equations). These equations are a system of quasilinear elliptic equations for Green’s functions $$\Phi _\omega (r)$$ and $$G_\omega (r) $$ and the pairing potential $$\Delta (r) $$ , which is determined from the equation of self-consistency by summation of the functions $$\Phi _\omega (r)$$ over the frequencies $$\omega $$ . To solve the quasilinear equations, we propose a special mixed finite element method, and to solve the self-consistency equations, we apply the successive approximation method and Anderson’s convergence acceleration algorithm. Results of calculations are provided for a structure with a wedge-shaped ferromagnetic layer.

Accelerating Benders decomposition for short-term hydropower maintenance scheduling

Maintenance of power generators is essential for reliable and efficient electricity production. Because generators under maintenance are typically inactive, optimal planning of maintenance activities must consider the impact of maintenance outages on the system operation. However, in hydropower systems finding a minimum cost maintenance schedule is a challenging optimization problem due to the uncertainty of the water inflows and the nonlinearity of the hydroelectricity production. Motivated by an industrial application problem, we formulate the hydropower maintenance scheduling problem as a two-stage stochastic program, and we implement a parallelized Benders decomposition algorithm for its solution. We obtain convex subproblems by approximating the hydroelectricity production using linear inequalities and indicator variables, which account for the nonlinear effect of the number of active generators in the solution. For speeding up the execution of our decomposition algorithm, we tailor and test seven techniques, including three new applications of special ordered sets, presolve and warm start for Benders acceleration. Given the large number of possible configurations of these acceleration techniques, we illustrate the application of statistical methods and computational experiments to identify the best performing configuration, which achieved a fourfold speedup of the decomposition algorithm. Results in an industrial setting confirm the high scalability on the number of scenarios of our parallelized Benders implementation.

Automatic detection of break-over phase onset in horses using hoof-mounted inertial measurement unit sensors.

A prolonged break-over phase might be an indication of a variety of musculoskeletal disorders and can be measured with optical motion capture (OMC) systems, inertial measurement units (IMUs) and force plates. The aim of this study was to present two algorithms for automatic detection of the break-over phase onset from the acceleration and angular velocity signals measured by hoof-mounted IMUs in walk and trot on a hard surface. The performance of these algorithms was evaluated by internal validation with an OMC system and a force plate separately. Seven Warmblood horses were equipped with two wireless IMUs which were attached to the lateral wall of the right front (RF) and hind (RH) hooves. Horses were walked and trotted over a force plate for internal validation while simultaneously the 3D position of three reflective markers, attached to lateral heel, lateral toe and lateral coronet of each hoof, were measured by six infrared cameras of an OMC system. The performance of the algorithms was evaluated by linear mixed model analysis. The acceleration algorithm was the most accurate with an accuracy between -9 and 23 ms and a precision around 24 ms (against OMC system), and an accuracy between -37 and 20 ms and a precision around 29 ms (against force plate), depending on gait and hoof. This algorithm seems promising for quantification of the break-over phase onset although the applicability for clinical purposes, such as lameness detection and evaluation of trimming and shoeing techniques, should be investigated more in-depth.

Efficient Deployment of Electric Vehicle Charging Infrastructure: Simultaneous Optimization of Charging Station Placement and Charging Pile Assignment

Charging infrastructure deployment is to seek the proper plan of settling charging stations and charging piles under multiple constraints, such as recharging demand, cruising range, etc., and it has been asserted as an NP-Complete problem. In this paper, we propose a multicriteria-oriented approach of efficiently deploying charging infrastructure to cope with the problem. We firstly formulate five realistic charging objective functions that exhibit a significant diminishing returns effect, i.e., submodularity, and then exploit the submodularity of these objectives to design the acceleration algorithms for Charging Station PLacement (CSPL) with the provable performance guarantees. The corresponding algorithms are respectively named Lazy Greedy with Direct Gain (LGDG) and Lazy Greedy with Effective Gain (LGEG), and they scale well to the road networks of arbitrary size. Relying on the inference that the linear combination of submodular functions is still a submodular function, we treat CSPL as a multicriteria optimization problem that can be efficiently solved by the proposed algorithms. Moreover, we employ Erlang-Loss system to gain an optimal Charging Pile ASsignment (CPAS), which is capable of reducing the gap between the growing complexity of charging demands and the constrained supply of charging resources in considering the correlation between the primary human activities and the charging process. The experimental evaluation with real data sets shows that, compared with the state-of-the-art methods, the proposed approach reveals better effectiveness and efficiency, and it offers a potent solution to the planning of charging infrastructure for electric vehicles with large-scale datasets in reality.

Acceleration scheme for particle transport in kinetic Monte Carlo methods.

Kinetic Monte Carlo (kMC) simulations are frequently used to study (electro-)chemical processes within science and engineering. kMC methods provide insight into the interplay of stochastic processes and can link atomistic material properties with macroscopic characteristics. Significant problems concerning the computational demand arise if processes with large time disparities are competing. Acceleration algorithms are required to make slow processes accessible. Especially, the accelerated superbasin kMC (AS-kMC) scheme has been frequently applied within chemical reaction networks. For larger systems, the computational overhead of the AS-kMC is significant as the computation of the superbasins is done during runtime and comes with the need for large databases. Here, we propose a novel acceleration scheme for diffusion and transport processes within kMC simulations. Critical superbasins are detected during the system initialization. Scaling factors for the critical rates within the superbasins, as well as a lower bound for the number of sightings, are derived. Our algorithm exceeds the AS-kMC in the required simulation time, which we demonstrate with a 1D-chain example. In addition, we apply the acceleration scheme to study the time-of-flight (TOF) of charge carriers within organic semiconductors. In this material class, time disparities arise due to a significant spread of transition rates. The acceleration scheme allows a significant acceleration up to a factor of 65 while keeping the error of the TOF values negligible. The computational overhead is negligible, as all superbasins only need to be computed once.

A Stride-Based Convolution Decomposition Method to Stretch CNN Acceleration Algorithms for Efficient and Flexible Hardware Implementation

To reduce multiplication operations in convolution of convolutional neural networks (CNNs), there are three widely used convolutional acceleration algorithms, i.e., Winograd, FFT and FFA. However, current accelerators based on these convolutional acceleration algorithms have issues on flexibility and efficiency. Firstly, some accelerators utilized a combination of these acceleration algorithms and employed multiple types of computational units to achieve their respective advantages. As a result, some computational units are left unused when the best-performing unit is working, which causes much area inefficiency. Secondly, current accelerators tend to choose small parameters of these convolutional acceleration algorithms to avoid unacceptable precision loss, as a result, they are hardly to support large kernel sizes and lack of flexibility. Thirdly, these acceleration algorithms are typically presented for 1-stride convolutions, consequently, few implementation considers the acceleration of large-stride convolutions, which is a major restriction to hardware flexibility. This paper proposed a stride-based convolution decomposition method (SCDM) to reform different convolution shapes (i.e., kernel sizes & strides) to an identical pattern. With the aid of SCDM, a Winograd-stretched and hardware-efficient design (WHD) is presented to utilize one uniform computational unit for the acceleration of different convolution shapes, which combines complementary performance advantages on both Winograd F(4,3) and F(4,2) units. Compared to current FFT-based or FFA-based works, WHD can stretch the use range of Winograd and simplify implementation, thereby achieving hardware flexibility and efficiency. Evaluation results show that 34.08%~55.41% operation reduction were achieved on six CNN models, while incurring a slight hardware overhead.

A corner smoothing method with feedrate blending for linear segments under geometric and kinematic constraints

To eliminate the disadvantages of linear interpolation and geometry-based smoothing methods, the one-step corner smoothing method with feedrate blending algorithm for linear segments is proposed in this article. In the proposed method, the variable acceleration and jerk optimization method will be first adopted to determine the dynamic performance of each linear segment for feedrate scheduling, which takes the different axial limits into consideration. Then, the feedrate scheduling method based on the S-type acceleration and deceleration algorithm is implemented to generate smooth feedrate profiles. Moreover, the feedrate blending algorithm is adopted to calculate the corner time analytically within the specified corner error and thus to generate the smooth trajectory without inserting a smooth curve at adjacent linear segments. With the presented one-step corner smoothing method, a smoother and more accurate trajectory could be generated with faster machining time. The results of simulations and machining experiments demonstrate the effectiveness of the proposed method.

Deep Neural Network Acceleration Based on Low-Rank Approximated Channel Pruning

Acceleration and compression on deep Convolutional Neural Networks (CNNs) have become a critical problem to develop intelligence on resource-constrained devices. Previous channel pruning can be easily deployed and accelerated without specialized hardware and software. However, weight-level redundancy is not well explored in channel pruning, which results in a relatively low compression ratio. In this work, we propose a Low-rank Approximated channel Pruning (LAP) framework to tackle this problem with two targeted steps. First, we utilize low-rank approximation to eliminate the redundancy within filter. This step achieves acceleration, especially in shallow layers, and also converts filters into smaller compact ones. Then, we apply channel pruning on the approximated network in a global way and obtain further benefits, especially in deep layers. In addition, we propose a spectral norm based indicator to coordinate these two steps better. Moreover, inspired by the integral idea adopted in video coding, we propose an evaluator based on Integral of Decay Curve (IDC) to judge the efficiency of various acceleration and compression algorithms. Ablation experiments and IDC evaluator prove that LAP can significantly improve channel pruning. To further demonstrate the hardware compatibility, the network produced by LAP obtains impressive speedup efficiency on the FPGA.

Continuous-time, high-cycle fatigue model: Validity range and computational acceleration for cyclic stress

The properties of the continuous-time, high-cycle fatigue model of Ottosen et al. (2008) is investigated for challenging stress states. We derive an analytical solution to the damage per cycle for cyclic, proportional stress. Numerical investigations for 7050-T7451 aluminum alloy and AISI 4340 steel alloy show exponential convergence to a constant damage per cycle for cyclic proportional stress, and Wynn’s epsilon algorithm for sequence acceleration improves the convergence rate. Fatigue damage is well predicted for an applied proportional stress, but damage is severely underpredicted for rotary stress states or combinations of tension/compression and torsion.

Mesoscale Particle Size Predictive Model for Operational Optimal Control of Bauxite Ore Grinding Process

This article investigates the use of a mesoscale kinetic model to cooperate with the operational optimal control of bauxite ore grinding process. In this article, we propose a new modeling framework where a discretized distributed parameter macroscale model and a mesoscale kinetic model are combined to predict the grinding product particle size. The mesoscale kinetic method does not need an explicit model of the process because it describes the process as a stochastic process. However, the high computational demand has prevented the kinetic model from using an online setting. We overcome this problem by embedding an acceleration algorithm based on the τ-leap method. The proposed model is validated using experimental data. Finally, a solution of the bauxite ore grinding operational optimal control is proposed and the cooperation of the predictive model with other modular is demonstrated.

A Novel GPU-Based Acceleration Algorithm for a Longwave Radiative Transfer Model

Graphics processing unit (GPU)-based computing for climate system models is a longstanding research area of interest. The rapid radiative transfer model for general circulation models (RRTMG), a popular atmospheric radiative transfer model, can calculate atmospheric radiative fluxes and heating rates. However, the RRTMG has a high calculation time, so it is urgent to study its GPU-based efficient acceleration algorithm to enable large-scale and long-term climatic simulations. To improve the calculative efficiency of radiation transfer, this paper proposes a GPU-based acceleration algorithm for the RRTMG longwave radiation scheme (RRTMG_LW). The algorithm concept is accelerating the RRTMG_LW in the g- p o i n t dimension. After implementing the algorithm in CUDA Fortran, the G-RRTMG_LW was developed. The experimental results indicated that the algorithm was effective. In the case without I/O transfer, the G-RRTMG_LW on one K40 GPU obtained a speedup of 30.98× over the baseline performance on one single Intel Xeon E5-2680 CPU core. When compared to its counterpart running on 10 CPU cores of an Intel Xeon E5-2680 v2, the G-RRTMG_LW on one K20 GPU in the case without I/O transfer achieved a speedup of 2.35×.

Multi-GPU, Multi-Node Algorithms for Acceleration of Image Reconstruction in 3D Electrical Capacitance Tomography in Heterogeneous Distributed System.

Electrical capacitance tomography (ECT) is one of non-invasive visualization techniques which can be used for industrial process monitoring. However, acquiring images trough 3D ECT often requires performing time consuming complex computations on large size matrices. Therefore, a new parallel approach for 3D ECT image reconstruction is proposed, which is based on application of multi-GPU, multi-node algorithms in heterogeneous distributed system. This solution allows to speed up the required data processing. Distributed measurement system with a new framework for parallel computing and a special plugin dedicated to ECT are presented in the paper. Computing system architecture and its main features are described. Both data distribution as well as transmission between the computing nodes are discussed. System performance was measured using LBP and the Landweber’s reconstruction algorithms which were implemented as a part of the ECT plugin. Application of the framework with a new network communication layer reduced data transfer times significantly and improved the overall system efficiency.

Process arrival pattern aware algorithms for acceleration of scatter and gather operations

Imbalanced process arrival patterns (PAPs) are ubiquitous in many parallel and distributed systems, especially in HPC ones. The collective operations, e.g. in MPI, are designed for equal process arrival times, and are not optimized for deviations in their appearance. We propose eight new PAP-aware algorithms for the scatter and gather operations. They are binomial or linear tree adaptations introducing additional process ordering and (in some cases) additional activities in a special background thread. The solution was implemented using one of the most popular open source MPI compliant library (OpenMPI), and evaluated in a typical HPC environment using a specially developed benchmark as well as a real application: FFT. The experimental results show a significant advantage of the proposed approach over the default OpenMPI implementation, showing good scalability and high performance with the FFT acceleration for the communication run time: 16.7% and for the total application execution time: 3.3%.

Optical Flow Based Co-located Reference Frame for Video Compression.

This paper proposes a novel bi-directional motion compensation framework that extracts existing motion information associated with the reference frames and interpolates an additional reference frame candidate that is co-located with the current frame. The approach generates a dense motion field by performing optical flow estimation, so as to capture complex motion between the reference frames without recourse to additional side information. The estimated optical flow is then complemented by transmission of offset motion vectors to correct for possible deviation from the linearity assumption in the interpolation. Various optimization schemes specifically tailored to the video coding framework are presented to further improve the performance. To accommodate applications where decoder complexity is a cardinal concern, a block-constrained speed-up algorithm is also proposed. Experimental results show that the main approach and optimization methods yield significant coding gains across a diverse set of video sequences. Further experiments focus on the trade-off between performance and complexity, and demonstrate that the proposed speed-up algorithm offers complexity reduction by a large factor while maintaining most of the performance gains.