Acceleration Algorithm Research Articles

A Novel Heuristic Emergency Path Planning Method Based on Vector Grid Map

Emergency path planning technology is one of the research hotspots of intelligent transportation systems. Due to the complexity of urban road networks and congested road conditions, emergency path planning is very difficult. Road congestion caused by urban emergencies directly affects the original road network structure. In this way, the static weight of the original road network is no longer suitable as the basis for path recommendation. To handle the dynamic situational road network, an equidistant grid emergency path planning framework will be designed. A novel situation grid road network model, based on situation information, is proposed and applied to an equidistant grid emergency path planning framework. A situational grid heuristic search will be proposed methodology based on this model, which can be used to detect the vehicles passing around the congestion area grid and the road to the destination in the shortest time. In the path planning methodology, a grid inspired search strategy based on quaternion function is included, which can make the algorithm converge to the target grid quickly. Three graph acceleration algorithms are proposed to improve the search efficiency of path planning algorithm. Finally, this paper will set up three experiments to verify our proposed method.

Accurate and efficient solution of electromagnetic scattering from randomly rough surface using MoM‐SMCG with adaptive quadrature

An adaptive quadrature method is proposed and implemented with the full-wave MoM-SMCG to accurately and efficiently solve the electromagnetic scattering from randomly rough surfaces. Without using high-order basis functions and dense surface discretisation, high numerical accuracy of computing the inner integrals for near-field impedance matrix elements is ensured by exploiting the orthogonal Legendre polynomials. Far-field interactions are calculated using the acceleration algorithms and the iterative solver to archive reasonable computational complexity and memory consumption. The proposed method is validated on solving the electromagnetic scattering and emission from rough ocean surfaces with fine-scale roughness and large dielectric constants. By comparing to the existing methods, simulation results indicate the proposed method has high computational efficiency and memory saving with also maintaining good accuracy.

Amata: An Annealing Mechanism for Adversarial Training Acceleration

Despite the empirical success in various domains, it has been revealed that deep neural networks are vulnerable to maliciously perturbed input data that much degrade their performance. This is known as adversarial attacks. To counter adversarial attacks, adversarial training formulated as a form of robust optimization has been demonstrated to be effective. However, conducting adversarial training brings much computational overhead compared with standard training. In order to reduce the computational cost, we propose an annealing mechanism, Amata, to reduce the overhead associated with adversarial training. The proposed Amata is provably convergent, well-motivated from the lens of optimal control theory and can be combined with existing acceleration methods to further enhance performance. It is demonstrated that on standard datasets, Amata can achieve similar or better robustness with around 1/3 to 1/2 the computational time compared with traditional methods. In addition, Amata can be incorporated into other adversarial training acceleration algorithms (e.g. YOPO, Free, Fast, and ATTA), which leads to further reduction in computational time on large-scale problems.

Binary Particle Swarm Optimization Algorithm for Kidney Exchanges Acceleration using Parallel MATLAB

Binary Particle Swarm Optimization Algorithm for Kidney Exchanges Acceleration using Parallel MATLAB

Development of MPS method and analytical approach for investigating RPV debris bed and lower head interaction in 1F Units-2 and 3

The onsite investigations of Fukushima Daiichi Nuclear Power Plant (1F) Units-2 and 3 indicate possibilities of multiple breaches of the Reactor Pressure Vessels (RPVs). In the meantime, some analytical works indicate possibilities that the core materials of 1F Units-2 and 3 were once relocated to the lower plenum of the RPVs and cooled (quenched) before the water inventory boiled-off (dry out) and the once-cooled debris re-melted. This study utilizes Lagrangian-based Moving Particle Semi-implicit (MPS) method to investigate such complex solid-liquid multiphase re-melting and the debris-vessel wall interactions to obtain new insight on 1F Units-2 and 3 RPV failure modes to help the future debris retrieval from these reactors. Two major modifications/developments have been carried out based on the previously developed MPS method. Namely, stabilization of rigid-body contact model and the further improvement of the speedup algorithm to enable large and long scale debris-bed re-melting analyses of the real plant scale. The numerical accuracy of the pressure boundary condition at fluid-wall boundary has also been improved. Sensitivity analyses have been carried out with the developed new MPS method. The results indicate the possibility that the lateral part of the RPV may have been subject to not only convective heat transfer of metallic melt pool, but also by conductive heat transfer by oxidic debris conglomerate. The results also indicate that following the initial vessel breach, discharge of metallic melt induces relocation of oxidic debris conglomerates, leading to concentration of the heat source around the central (bottom) part of the lower head. However, large uncertainties associated with 1F are acknowledged and further model validations are necessary before drawing further insights.

Acceleration Design for Continuum Topology Optimization by Using Pix2pix Neural Network

Aiming at speeding up the process of topology optimization and generating a novel high-quality topology configuration, a topology optimization design with lightweight structure based on deep learning is investigated. First, Independent Continuous Mapping (ICM) method is employed to establish the original configurations including the intermediate configurations and the corresponding final configurations, and the original configurations are padded and merged to form the dataset required by the deep learning network. Then the high-dimensional mapping relationship between the intermediate configuration and the final configuration is created by using Pix2pix neural network (Pix2pix NN), which transforms the topology optimization into image-to-image translation problem. Finally, the acceleration algorithm is utilized to accelerate the iteration process by the pre-trained network. Numerical examples show that the coupling method is feasible and efficient in topology optimization design. The method provides a new solution for topology optimization design to shorten the iteration process and broaden the application of ICM method.

Quantum-enhanced algorithms for classical target detection in complex environments

Quantum computational approaches to some classic target identification and localization algorithms, especially for radar images, are investigated, and are found to raise a number of quantum statistics and quantum measurement issues with much broader applicability. Such algorithms are computationally intensive, involving coherent processing of large sensor data sets in order to extract a small number of low profile targets from a cluttered background. Target enhancement is accomplished through accurate statistical characterization of the environment, followed by optimal identification of statistical outliers. The key result of the work is that the environmental covariance matrix estimation and manipulation at the heart of the statistical analysis actually enables a highly efficient quantum implementation. The algorithm is inspired by recent approaches to quantum machine learning, but requires significant extensions, including previously overlooked `quantum analog--digital' conversion steps (which are found to substantially increase the required number of qubits), `quantum statistical' generalization of the classic phase estimation and Grover search algorithms, and careful consideration of projected measurement operations. Application regimes where quantum efficiencies could enable significant overall algorithm speedup are identified. Key possible bottlenecks, such as data loading and conversion, are identified as well.

GPU coprocessors as a service for deep learning inference in high energy physics

In the next decade, the demands for computing in large scientific experiments are expected to grow tremendously. During the same time period, CPU performance increases will be limited. At the CERN Large Hadron Collider (LHC), these two issues will confront one another as the collider is upgraded for high luminosity running. Alternative processors such as graphics processing units (GPUs) can resolve this confrontation provided that algorithms can be sufficiently accelerated. In many cases, algorithmic speedups are found to be largest through the adoption of deep learning algorithms. We present a comprehensive exploration of the use of GPU-based hardware acceleration for deep learning inference within the data reconstruction workflow of high energy physics. We present several realistic examples and discuss a strategy for the seamless integration of coprocessors so that the LHC can maintain, if not exceed, its current performance throughout its running.

Generalizations of Shanks transformation and corresponding convergence acceleration algorithms via pfaffians

Firstly, a new sequence transformation that can be expressed in terms of a ratio of two pfaffians is derived based on a special kernel. It can be regarded as a direct generalization of Aitken’s Δ2 process from the point of view of pfaffians and then the corresponding convergence acceleration algorithm is constructed. Numerical examples with applications of this algorithm are also presented. Secondly, we find a way to generalize the Shanks transformation via pfaffians so that a larger class of new sequence transformations are derived. The corresponding recursive algorithms are also proposed.

Whole Brain Myelin Water Mapping in One Minute Using Tensor Dictionary Learning With Low-Rank Plus Sparse Regularization.

The quantification of myelin water content in the brain can be obtained by the multi-echo [Formula: see text] weighted images ( [Formula: see text]WIs). To accelerate the long acquisition, a novel tensor dictionary learning algorithm with low-rank and sparse regularization (TDLLS) is proposed to reconstruct the [Formula: see text]WIs from the undersampled data. The proposed algorithm explores the local and nonlocal similarity and the global temporal redundancy in the real and imaginary parts of the complex relaxation signals. The joint application of the low-rank constraints on the dictionaries and the sparse constraints on the core coefficient tensors improves the performance of the tensor-based recovery. Parallel imaging is incorporated into the TDLLS algorithm (pTDLLS) for further acceleration. A pulse sequence is proposed to prospectively undersample the Ky-t space to obtain the whole brain high-quality myelin water fraction (MWF) maps within 1 minute at an undersampling rate (R) of 6.

An Adaptive Iterated Greedy algorithm for distributed mixed no-idle permutation flowshop scheduling problems

Distributed flow shop scheduling is a very interesting research topic. This paper studies the distributed permutation flow shop scheduling problem with mixed no-idle constraints, which have important applications in practice. The optimization goal is to minimize total flowtime. A mixed-integer linear programming model is presented and an Adaptive Iterated Greedy (AIG) algorithm with the sample length changing according to the search process is designed. A restart strategy is also introduced to escape from local optima. Additionally, to further improve the performance of the algorithm, swap-based local search methods and acceleration algorithms for swap neighborhoods are proposed. Referenced Local Search (RLS), which shows better performance in solving scheduling problems, is also used in our algorithm. In the destruction stage, the job to be removed is selected according to the degree of influence on the total flowtime. In the initialization and construction phase, when a job is inserted, the jobs before and after the insertion position are removed and re-inserted into a better position to improve the algorithm search performance. A detailed design experiment is carried out to determine the best parameter configuration. Finally, large-scale experiments show that the proposed AIG algorithm is the best-performing one among all the algorithms in comparison.

CNN-LSTM Learning Approach-Based Complexity Reduction for High-Efficiency Video Coding Standard

High-Efficiency Video Coding provides a better compression ratio compared to earlier standard, H.264/Advanced Video Coding. In fact, HEVC saves 50% bit rate compared to H.264/AVC for the same subjective quality. This improvement is notably obtained through the hierarchical quadtree structured Coding Unit. However, the computational complexity significantly increases due to the full search Rate-Distortion Optimization, which allows reaching the optimal Coding Tree Unit partition. Despite the many speedup algorithms developed in the literature, the HEVC encoding complexity still remains a crucial problem in video coding field. Towards this goal, we propose in this paper a deep learning model-based fast mode decision algorithm for HEVC intermode. Firstly, we provide a deep insight overview of the proposed CNN-LSTM, which plays a kernel and pivotal role in this contribution, thus predicting the CU splitting and reducing the HEVC encoding complexity. Secondly, a large training and inference dataset for HEVC intercoding was investigated to train and test the proposed deep framework. Based on this framework, the temporal correlation of the CU partition for each video frame is solved by the LSTM network. Numerical results prove that the proposed CNN-LSTM scheme reduces the encoding complexity by 58.60% with an increase in the BD rate of 1.78% and a decrease in the BD-PSNR of -0.053 dB. Compared to the related works, the proposed scheme has achieved a best compromise between RD performance and complexity reduction, as proven by experimental results.

Local search-based metaheuristics for the robust distributed permutation flowshop problem

Distributed Permutation Flowshop Scheduling Problem (DPFSP) has become a hot issue in recent years. However, DPFSP with uncertain processing times (DPFSP_UPT) has not been addressed so far although real manufacturing systems often encounter various uncertain factors which make processing times uncertain. In this paper, two local-search based metaheuristics, i.e., an Iterated Greedy algorithm (sIG) and an Iterated Local Search algorithm (sILS), are presented to solve the DPFSP_UPT with makespan criterion. A robust model of expect-risk rule is used to describe the DPFSP_UPT. A valid heuristic based on the well-known NEH heuristic is proposed to initialize sIG and sILS. An acceleration algorithm is adapted to save computational efforts. A destruction procedure with dynamic sizes is presented to enhance the exploration capability for sIG. A feature with the characteristic that small changes in the solution can preserve the properties of good solutions in the iterative process is applied by sILS. Both sIG and sILS use multiple local search methods to produce promising solutions by exploiting diverse search areas. Extensive experiments show that the proposed sIG and sILS perform significantly better than the five competing algorithms adapted in the literature, but there is no significant difference between sIG and sILS and each has its own advantages for different instances.

Near-Optimal Graph Signal Sampling by Pareto Optimization.

In this paper, we focus on the bandlimited graph signal sampling problem. To sample graph signals, we need to find small-sized subset of nodes with the minimal optimal reconstruction error. We formulate this problem as a subset selection problem, and propose an efficient Pareto Optimization for Graph Signal Sampling (POGSS) algorithm. Since the evaluation of the objective function is very time-consuming, a novel acceleration algorithm is proposed in this paper as well, which accelerates the evaluation of any solution. Theoretical analysis shows that POGSS finds the desired solution in quadratic time while guaranteeing nearly the best known approximation bound. Empirical studies on both Erdos-Renyi graphs and Gaussian graphs demonstrate that our method outperforms the state-of-the-art greedy algorithms.

An algorithm for generation of RVEs of composites with high particle volume fractions

This work proposes a new algorithm entitled “sequential absorption algorithm” to generate Representative Volume Elements (RVEs) of randomly distributed spherical particles reinforced composites by combining the Random Sequential Absorption (RSA) algorithm and the molecular dynamics based method. The proposed algorithm overcomes limitations of the RSA algorithm and is capable of efficiently generating the RVEs with the high Particle Volume Fractions (PVFs) (≥50.0%). The proposed algorithm comprises of the modified RSA algorithm, the detection algorithms of the collisions between particles and between a particle and matrix surface(s), the post-collision particle velocity update algorithm, the particle periodic image generation algorithm and the acceleration algorithm. Particle distribution in the generated RVEs is analyzed through several statistical functions and consistent with the completely spatial random pattern. Regarding the elastic properties of composites, the critical sizes of RVEs and meshed elements are determined. The agreement of the elastic properties of composites with the different PVFs acquired using the finite element homogenization method, the experimental tests and the analytical models exhibits the validation of the proposed algorithm to generate RVEs of composites.

Experimental Study on the Optimal Strategy for Power Regulation of Governing System of Hydropower Station

Active power instability during the power regulation process is a problem that affects the operation security of hydropower stations and the power grid. This paper focuses on the dynamic response to power regulation of a hydro-turbine governor in the power control mode. Firstly, the mathematical model for the hydro-turbine governing system connected to the power grid is established. Then, considering the effect of water hammer and the guide vane operating speed on power oscillation and reverse power regulation, a novel control strategy based on the S-curve acceleration and deceleration control algorithm (S-curve control algorithm) is proposed to improve power regulation. Furthermore, we carried out field tests in a real hydropower station in order to compare the regulation quality of the novel control strategy based on the S-curve control algorithm with the traditional linear control strategy. Finally, the obtained results show that the proposed optimal control strategy for the hydro-turbine governor improves the stability of power regulation by effectively suppressing reverse power regulation and overshoot. This study provides a good solution for the instability of power and reverse power regulation during the regulation process of the hydro-turbine governor in the power control mode.

Modeling Acceleration Properties for Flexible INTRA HEVC Complexity Control

It is a very well-known fact, that the high complexity of the High Efficiency Video Coding standard (HEVC) is the main hurdle for its wide deployment and use. To tackle this problem, a number of recent research outcomes exploit heuristic algorithms and machine learning, including deep learning, to reduce the coding complexity. However, in most cases, each encoder module, i.e., encoding process, is first accelerated individually, and then different acceleration algorithms are manually combined. Without a holistic strategy, the acceleration potential of multi-module combination is not exploited and the Rate-Distortion (RD) loss is generally not well controlled. To tackle these shortcomings, this paper exploits the acceleration properties of different modules, i.e., the numerical representation of potential time saving and possible RD loss, from which a heuristic model is explored. Then a Heuristic Model Oriented Framework (HMOF) is proposed which adapts the properties of modules to underlying acceleration algorithms. In the framework, two advanced acceleration algorithms, including Border Considered CNN (BC-CNN)-based Coding Unit (CU) partition and Naive Bayes-based Prediction Unit (PU) partition, are proposed for the CU and PU modules, respectively. Further, by leveraging the heuristic model as the guidance to combine the proposed acceleration algorithms, HMOF is globally optimized, where different time saving budgets are wisely allocated to different modules and a theoretically minimal RD loss is achieved. According to the experimental results, through fusing a suitable deep learning technique and a Bayes-Based prediction, the proposed acceleration framework HMOF enable multiple acceleration choices. Here the proposed joint optimization strategy help to make a choice leading to the best cost-performance. Furthermore, within the proposed framework, intra coding time can be precisely controlled with negligible Bjøntegaard delta bit-rate (BDBR) loss. In this context, as a complexity control method, HMOF outperforms the state-of-the-art complexity reduction algorithms under a similar complexity reduction ratio. These results partially demonstrate the superiority of the proposed technique.

A dynamic network traffic classifier using supervised ML for a Docker-based SDN network

With the rapid technological growth in the last decades, the number of devices and users has drastically increased. Software-defined networking (SDN) with machine learning (ML) has become an emerging solution for network scheduling, quality of service (QoS), resource allocations, and security. This paper focuses on the implementation of a network traffic classifier using a novel Docker-based SDN network. ML offers good performance to real-time traffic solutions without depending on well-known TCP or UDP port numbers, IP addresses, or encrypted payloads. In this paper, using three ML techniques, we first classify network flows with 3, 5, and 7 parameters giving up to 97.14% accuracy. Additionally, we present a new performance accelerator algorithm (PAA), which incorporates these three ML classifiers and accelerates the overall performance significantly. We then propose a dynamic network classifier (DNC) generated from PAA over a novel Docker-based SDN network. Finally, we propose a new controller algorithm for Ryu platforms, which integrates the DNC and classifies both TCP and UDP flows in real-time. Based on the evaluations, an improvement in latency performance has been demonstrated, where analysing a packet, controller processing time takes on an average of 10 µs. This study will certainly serve to further research on optimising SDN and QoS.

An advanced method for efficiently generating composite RVEs with specified particle orientation

In this work, an improved version of the Random Sequential Absorption (RSA) algorithm, comprising of the particle re-orientation algorithm, the particle intersection checking algorithm, the particle periodicity constraint algorithm and the acceleration algorithm, is proposed to efficiently generate Representative Volume Elements (RVEs) of spheroidal particles reinforced composites with specified particle orientations. The re-orientation of the particles is realized using the gradient descent based optimization method, and the RSA algorithm is accelerated by combining with the RVE subcell method and the bounding sphere concept. Several statistical functions are introduced to analyze the distributions of the orientation and centroids of the particles in the RVEs generated by the improved algorithm. The results show that the particle orientations of the generated RVEs match well with the specified particle orientations, and the centroids of the particles in the generated RVEs are not completely randomly distributed. Based on the generated RVEs, the elastic properties and coefficients of thermal expansion of spheroidal particles reinforced composites are predicted by using the FE homogenization method, and the predicted thermo-elastic properties of the composites agree well with those of the analytical models. The advantage of the improved algorithm lies in two aspects: (1) much better computational efficiency and (2) capability of generating the RVEs with specified particle orientations.

Mixed Finite Element-Second Order Upwind Fractional Step Difference Scheme of Darcy–Forchheimer Miscible Displacement and Its Numerical Analysis

The incompressible miscible displacement of three-dimensional Darcy–Forchheimer flow is discussed in this paper, and the mathematical model is formulated by two partial differential equations, a Darcy–Forchheimer flow equation for the pressure and a convection-diffusion equation for the concentration. The pressure plays an important role and determines the Darcy–Forchheimer velocity. A conservative mixed finite element is used to approximate the pressure and the velocity, and the computational accuracy of Darcy–Forchheimer velocity is improved by one order. A second-order upwind fractional step difference scheme is adopted to obtain the concentration, and numerical oscillation and dispersion are eliminated. Computational work is reduced greatly by decomposing the whole computation into successive one-dimensional subcomputations, where the speedup solvers are used. Applying energy-norm method, operator decomposition, speedup algorithm and induction hypotheses, we derive an optimal second-order estimates in $$L^2$$ norm. Numerical experiments are given to show the theoretical accuracy and the applicability for solving actual problems.