Acceleration Method Research Articles

Unlocking the potential of edge computing for hyperspectral image classification: An efficient low-energy strategy

Despite recent improvements, the computing capability of Edge Computing devices is still inferior to high-end servers, so special methodologies are required to consider the computing environment while developing algorithms. In the present work, we propose a hybrid technique to make the classification of Hyperspectral Images feasible and effective through a Convolutional Neural Network on low-power and high-performance sensor devices. More specifically, we combine two strategies: we initially use the Principal Component Analysis method to discard non-significant wavelengths and shrink the dataset; then, we apply a process acceleration method to boost performance by implementing a form of GPU-based parallelism. The experiments demonstrate the technique’s effectiveness in terms of performance and energy consumption: it enables correct classifications even with low-power devices often deployed on Unmanned Aerial Vehicles, where the network connection is unpredictable or erratic.

Anti-sway control of variable rope length container crane based on phase plane trajectory planning

An overhead crane system is a complex time-varying system if it includes simultaneous hoisting and moving operations. To address the problem that time-varying systems usually have no exact analytical solution, an anti-swing method based on linear approximation of the phase plane trajectory is proposed. Specifically, the phase plane trajectory of the swing angle is first analyzed and it is found that the double-step acceleration method with unequal acceleration amplitude and acceleration time has a combination of limit loops in the phase plane trajectory of the swing angle when the rope length varies, so the average rope length of each section is used to approximate the natural oscillation frequency. Next, the input shaping technique is used to shape the input signal into the desired output signal so that the load swing angle can follow the planned trajectory. The desired acceleration amplitude and switching time are then derived based on the geometric relationship between the swing angle in the phase plane and the specific physical constraints. Finally, the feasibility of the proposed method is verified by numerical simulations.

Approximation of the Tikhonov regularization parameter through Aitken's extrapolation

In the present work, we study the determination of the regularization parameter and the computation of the regularized solution in Tikhonov regularization, by the Aitken's extrapolation method. In particular, this convergence acceleration method is adjusted for the approximation of quadratic forms that appear in regularization methods, such as the generalized cross-validation method, the quasi-optimality criterion, the Gfrerer/Raus method and the Morozov's discrepancy principle. We present several numerical examples to illustrate the effectiveness of the derived estimates for approximating the regularization parameter for several linear discrete ill-posed problems and we compare the described method with further existing methods, for the determination of the regularized solution.

Topology Optimization of Self-supporting Porous Structures Based on Triply Periodic Minimal Surfaces

With the advent of additive manufacturing (AM), topology optimization (TO) has become increasingly important in structural design. A key component of structural design is generating the interior shape of an object under predefined force conditions and specific constraints. Fabricating such models by layer-based AM suffers from the problem of adding and removing interior supporting structures, which can produce artifacts in the final product. In this paper, we propose an algorithm to design a self-supporting porous structure by integrating overhang constraints into the topology optimization framework. A minimum thickness constraint is also built into the optimization process to automatically enforce printability of the resulting structures. We utilize triply periodic minimal surfaces (TPMSs) to represent interior structures which can be analyzed, optimized and stored directly using analytical functions. We apply several acceleration methods including super element strategy, multigrid algorithm and GPU-based solver to handle models comprising of several million of elements efficiently. Numerical results indicate that the optimized interior structures obtained by our approach exhibit improved printability as they largely satisfy manufacturing requirements on overhang angles and minimal thicknesses.

The Prediction of Shelf Life of Carrot Chips by Using Acceleration Method

Food processing technology to produce meals that ready to eat is one alternative way to reduce damages of horticulture product. To ensure the carrot chips are still suitable for consumption, the information of the shelf life of this product is needed. Method to determine shelf life of carrot chips is by using Accelerated Storage Studies (ASS) with Arhenius approach using 7°C, 15°C, 25°C and 30°C as acceleration condition. In an empirical research, carrot chips were packaged using aluminum foil with thickness of 50 μm, 80 μm and 100 μm and stored at 30°C, 37°C and 45°C. The products were evaluated every weeks for eight weeks of storage. Parameters observed in both studies were moisture content, crispness and free fatty acids (FFA). Result showed that aluminum foil with thickness 100 μm maintained quality of carrot chips during storage. The product packaged in aluminum foil thickness of 100 μm had the longest shelf life in terms of FFA parameter using the acceleration method reaching 66.99 to 103.21 days.

Battery State of Health Model Estimation Based on Variable Projection Algorithm Using Steffensen Acceleration Method

In this paper, the variable projection based Steffensen acceleration (VP-SA) algorithm is proposed for battery state of health (SOH) estimation. The VP-SA algorithm exploits the separability of variables to reduce the dimensionality of the parameters, which can improve the estimation efficiency. In addition, based on the Steffensen acceleration method, the convergence rates of the VP algorithm can be increased. The simulation example shows the effectiveness of the proposed algorithm.

Evaluasi Time Schedule Menggunakan Network Planning (Studi Kasus: Pada Proyek Pembangunan Gedung Rawat Inap Kelas III Tahap III Rumah Sakit Umum Daerah Bangkinang, Riau)

Time Schedule is a detailed division of work duration on each project work. The problem in this study is the ineffectiveness of the duration of the work and the amount of work done is not in accordance with the planned duration. The purpose of the study was to evaluate the time schedule using the network planning method to determine the critical trajectory by using Microsoft Project 2016 software and to accelerate the duration of work from its original duration by using the acceleration method of adding labor, increasing working hours and accelerating the work shift method. The conclusion of time schedule evaluation using network planning is obtained as many as 50 work items which are included in the critical path. After accelerating with three acceleration methods, it was found that the results of accelerating the addition of manpower which cut the time by 3 days from the original duration, acceleration by adding 2 hours of work cut 7 working days and acceleration with the work shift method which cut the duration to 21 working days from the original duration. which is 206 working days, then of the three methods of acceleration the most effective is the shift method.

An accelerating convolutional neural networks via a 2D entropy based-adaptive filter search method for image recognition

The success of CNNs for various vision tasks has been accompanied by a significant increase in required FLOPs and parameter quantities, which has impeded the deployment of CNNs on devices with limited computing resources and power budgets. Network pruning, which compresses and accelerates CNN models, is an effective solution to this issue. Some studies have considered pruning as a special case of neural network search (NAS) in recent years. However, existing techniques are often computationally complex or prone to sub-optimal pruning results. As such, this paper proposes a novel acceleration method via a 2D Entropy based-Adaptive Filter Search (2EAFS). The importance of corresponding filters, measured by utilizing the amount of information contained in feature maps, is employed as a theoretical guide to simplify the complex exhaustive search process. Information entropy is then normalized layer by layer and the resulting value is used to calculate a layer-wise importance score in a single step. Additionally, a sparse constraint equation is constructed based on the negative correlation between filter pruning rates and the importance of convolutional layers. The Nelder–Mead search algorithm is then adopted to quickly and adaptively determine the optimal pruning architecture. Finally, importance weights are inherited using the pruning rate and 2D entropy and model performance are restored through fine-tuning. Extensive experiments conducted with the CIFAR-10/100, ILSVRC-2012, NWPU-RESISC45 and CUB-200-2011 datasets showed this approach achieved considerable accuracy increases, with significant reductions in FLOPs and required parameters that surpassed current state-of-the-art methods by a wide margin. For example, 2EAFS achieved a 44.1% reduction in FLOPs over ResNet-50, with only a 0.53% Top-5 accuracy decrease for ILSVRC-2012.

A Review of Artificial Intelligence in Embedded Systems.

Advancements in artificial intelligence algorithms and models, along with embedded device support, have resulted in the issue of high energy consumption and poor compatibility when deploying artificial intelligence models and networks on embedded devices becoming solvable. In response to these problems, this paper introduces three aspects of methods and applications for deploying artificial intelligence technologies on embedded devices, including artificial intelligence algorithms and models on resource-constrained hardware, acceleration methods for embedded devices, neural network compression, and current application models of embedded AI. This paper compares relevant literature, highlights the strengths and weaknesses, and concludes with future directions for embedded AI and a summary of the article.

Undersampled Diffusion-Weighted 129Xe MRI Morphometry of Airspace Enlargement: Feasibility in Chronic Obstructive Pulmonary Disease

Multi-b diffusion-weighted hyperpolarized gas MRI measures pulmonary airspace enlargement using apparent diffusion coefficients (ADC) and mean linear intercepts (Lm). Rapid single-breath acquisitions may facilitate clinical translation, and, hence, we aimed to develop single-breath three-dimensional multi-b diffusion-weighted 129Xe MRI using k-space undersampling. We evaluated multi-b (0, 12, 20, 30 s/cm2) diffusion-weighted 129Xe ADC/morphometry estimates using a fully sampled and retrospectively undersampled k-space with two acceleration-factors (AF = 2 and 3) in never-smokers and ex-smokers with chronic obstructive pulmonary disease (COPD) or alpha-one anti-trypsin deficiency (AATD). For the three sampling cases, mean ADC/Lm values were not significantly different (all p > 0.5); ADC/Lm values were significantly different for the COPD subgroup (0.08 cm2s-1/580 µm, AF = 3; all p < 0.001) as compared to never-smokers (0.05 cm2s-1/300 µm, AF = 3). For never-smokers, mean differences of 7%/7% and 10%/7% were observed between fully sampled and retrospectively undersampled (AF = 2/AF = 3) ADC and Lm values, respectively. For the COPD subgroup, mean differences of 3%/4% and 11%/10% were observed between fully sampled and retrospectively undersampled (AF = 2/AF = 3) ADC and Lm, respectively. There was no relationship between acceleration factor with ADC or Lm (p = 0.9); voxel-wise ADC/Lm measured using AF = 2 and AF = 3 were significantly and strongly related to fully-sampled values (all p < 0.0001). Multi-b diffusion-weighted 129Xe MRI is feasible using two different acceleration methods to measure pulmonary airspace enlargement using Lm and ADC in COPD participants and never-smokers.

Graphene Mediates Charge Transfer between Lead Chromate and a Cobalt Cubane Cocatalyst for Photocatalytic Water Oxidation.

The interfacial barrier of charge transfer from semiconductors to cocatalysts means that the photogenerated charges cannot be fully utilized, especially for the challenging water oxidation reaction. Using cobalt cubane molecules (Co4 O4 ) as water oxidation cocatalysts, we rationally assembled partially oxidized graphene (pGO), acting as a charge-transfer mediator, on the hole-accumulating {-101} facets of lead chromate (PbCrO4 ) crystal. The assembled pGO enables preferable immobilization of Co4 O4 molecules on the {-101} facets of the PbCrO4 crystal, which is favorable for the photogenerated holes transferring from PbCrO4 to Co4 O4 molecules. The surface charge-transfer efficiency of PbCrO4 was boosted by selective assembly of pGO between PbCrO4 and Co4 O4 molecules. An apparent quantum efficiency for photocatalytic water oxidation on the Co4 O4 /pGO/PbCrO4 photocatalyst exceeded 10 % at 500 nm. This strategy of rationally assembling charge-transfer mediator provides a feasible method for acceleration of charge transfer and utilization in semiconductor photocatalysis.

Graphene Mediates Charge Transfer between Lead Chromate and a Cobalt Cubane Cocatalyst for Photocatalytic Water Oxidation

AbstractThe interfacial barrier of charge transfer from semiconductors to cocatalysts means that the photogenerated charges cannot be fully utilized, especially for the challenging water oxidation reaction. Using cobalt cubane molecules (Co4O4) as water oxidation cocatalysts, we rationally assembled partially oxidized graphene (pGO), acting as a charge‐transfer mediator, on the hole‐accumulating {−101} facets of lead chromate (PbCrO4) crystal. The assembled pGO enables preferable immobilization of Co4O4 molecules on the {−101} facets of the PbCrO4 crystal, which is favorable for the photogenerated holes transferring from PbCrO4 to Co4O4 molecules. The surface charge‐transfer efficiency of PbCrO4 was boosted by selective assembly of pGO between PbCrO4 and Co4O4 molecules. An apparent quantum efficiency for photocatalytic water oxidation on the Co4O4/pGO/PbCrO4 photocatalyst exceeded 10 % at 500 nm. This strategy of rationally assembling charge‐transfer mediator provides a feasible method for acceleration of charge transfer and utilization in semiconductor photocatalysis.

Heavy-ball-based hard thresholding algorithms for sparse signal recovery

The hard thresholding technique plays a vital role in the development of algorithms for sparse signal recovery. By merging this technique and heavy-ball acceleration method which is a multi-step extension of the traditional gradient descent method, we propose the so-called heavy-ball-based hard thresholding (HBHT) and heavy-ball-based hard thresholding pursuit (HBHTP) algorithms for signal recovery. It turns out that the HBHT and HBHTP can successfully recover a k-sparse signal if the restricted isometry constant of the measurement matrix satisfies δ3k<0.618 and δ3k<0.577, respectively. The guaranteed success of HBHT and HBHTP is also shown under the conditions δ2k<0.356 and δ2k<0.377, respectively. Moreover, the finite convergence of HBHTP and stability of the two algorithms are also established in this paper. Simulations on random problem instances are performed to compare the performance of the proposed algorithms and several existing ones. Empirical results indicate that the HBHTP performs very comparably to a few existing algorithms and it takes less average time to achieve the signal recovery than these existing methods.

Graph neural networks and implicit neural representation for near-optimal topology prediction over irregular design domains

This paper proposes a deep neural network-based topology optimization acceleration method for irregular design domains that predicts (near-)optimal topologies. A topology optimization problem is a complex non-Euclidean data, which can be embedded in a graph form, and a graph neural network encodes it to Euclidean data such as vectors and matrices. The encoded information is applied to a multi-layer perceptron-based implicit neural representation model, and the multi-layer perceptron approximator predicts the compliance optimal material distribution. The prediction performance of the proposed encoder-approximator architecture is evaluated for several topology optimization problems. The trained network provides 96.6% compliance accuracy, except for 8.0% of the outliers. The two criteria have been investigated to estimate potential outliers, and post-optimization can resolve the outlier within fewer iterations than the original optimization.

A robust reliability evaluation model with sequential acceleration method for power systems considering renewable energy temporal-spatial correlation

As the penetration of renewable energy increases, it becomes critical to evaluate the power system reliability, while considering the correlations among renewable energy sources (RESs) and their uncertainties. In this paper, the robust reliability evaluation model based on the minimum volume enclosing ellipsoid (MVEE) algorithm is established, and the sequential acceleration method is proposed to improve the convergence. First, the robust reliability evaluation model is built, where the robust state analysis is performed under each sampling. Second, the sequential Cross-Entropy-Latin Hypercube Sampling (CE-LHS) acceleration method is presented, which first obtains the optimal probability distribution of system parameters, and then forces the sampling values to be evenly distributed to overcome the truncated tail effect. Correspondingly, the reliability indices are improved. Third, for state analysis, the robust multi-period optimal load shedding (RMPOLS) model is established, where the dynamic performance of energy storage and system ramping rate constraints are described in detail. Therefore, the periods are strongly coupled, and the impact of RES uncertainties is carefully considered. Besides, to address the temporal-spatial correlation of RESs, the MVEE algorithm is introduced to generate the convex hull of RESs scenarios. Therefore, the RMPOLS model can be transformed into a second-order cone form. Finally, to further reduce the computational complexity, the branch flow constraints scanning strategy is proposed, which can quickly remove inactive constraints in advance. Numerical results prove that the MVEE-based robust reliability evaluation model considers the temporal-spatial correlation of RES and thus improves the system reliability by almost 30 %. Besides, branch flow constraint scanning can eliminate more than 65 % of the SOC constraints. Moreover, CE-LHS can improve the convergence by about 8 times compared with Monte Carlo Sampling (MCS) while the error is less than 1 %.

Efficient Spectral Graph Convolutional Network Deployment on Memristive Crossbars

Graph Neural Networks (GNNs) have attracted increasing research interest for their remarkable capability to model graph-structured knowledge. However, GNNs suffer from intensive data exchange and poor data locality, which will cause critical performance and energy bottlenecks under conventional complementary metal oxide semiconductor (CMOS)-based von-Neumann computing architectures (graphics processing unit (GPU), central processing unit (CPU)) for the “Memory Wall” issue. Fortunately, memristive crossbar-based computation has emerged as one of the most promising neuromorphic computing architectures, which has been widely studied as the computing platform for convolutional neural network (CNNs), recurrent neural network (RNNs), spiking neural network (SNNs), etc. This paper proposes the deployment of spectral graph convolutional networks (GCNs) on memristive crossbars. Further, based on the structure of GCNs (extremely high sparsity and unbalanced non-zero data distribution) and the neuromorphic characteristics of memristive crossbar circuit, we propose the acceleration method that consists of Sparse Laplace Matrix Reordering and Diagonal Block Matrix Multiplication. The simulated experiment on memristor crossbars achieves 90.3% overall accuracy on the supervised learning graph dataset (QM7), and compared with the original computation, the proposed acceleration computing framework (with half-size diagonal blocks) achieves a 27.3% reduction of memristor number. Additionally, on the unsupervised learning dataset (Karate club), our method shows no loss of accuracy with half-size diagonal block mapping and reaches a 32.2% reduction of memristor number.

Substantiation of the Structural and Technological Parameters of the Working Body of the Milling Grain Shredder

Introduction. One of the main components in the diet of farm animals is concentrated feed, which in Russia is grinded by shredders. At the same time, in the final product, there is a high dust content, which can be reduced by using grinders with a different principle of grinding. The design of a milling shredder, in which grains are crushed by cutting, is proposed. Aim of the Article. The article presents theoretical studies of the influence of design and technological parameters of a milling shredder on the conditions of grain motion along the tooth of the working body. Materials and Methods. The theoretical study of the developed shredder was carried out using averaged acceleration methods. The tooth surface of the working organ was studied. Results. As a result of theoretical studies, equations describing the motion of a particle along the working body tooth surface were obtained. There has been developed a program to facilitate the process of analyzing the influence of the structural and technological parameters of the shredder on the process of particle motion along the tooth surface. Discussion and Conclusion. There is a directly proportional decrease in the time of motion of the particle along the working body tooth surface with increasing the angular velocity. At the same time, increasing the angle leads to the increase of the time of the particle along the tooth surface. The increase of angular velocity values leads to decreasing particle displacement along the axis. At the angle of 45° displacement along the axis is 1.1‒1.5 times less than at an angle of 70°. At high angular velocities, the moving along the axis is minimal. The analysis, given the need to create a smoother motion of the particle along the helical line, shows that the most rational will be the following values: angle 70° and angular velocity of the working body 185‒206 rad/s.

Assessment of contact interface on seismic response of tunnels by the modified response acceleration method

The contact interface of the tunnel with composite lining will affect the seismic response of the tunnel. It is necessary to study the effect of interface properties on the seismic response of the tunnel, including the tunnel with buffer layer and without buffer layer. Initially, the establishment of the finite element (FE) model is introduced in detail, including the convergence-confinement method and the tracking element method. Then, a modified response acceleration method (MRAM) is proposed and verified, and it is proved to be accurate and convenient. Finally, based on the MRAM, the influence of the contact interface on the seismic response of the tunnel is studied, including the tunnel with buffer layer and without buffer layer. At the same time, different PGA and buffer layer thickness are considered. The results show that the property of the contact interface significantly affect the seismic response of the tunnel, including the internal force and tensile damage of the tunnel. Moreover, PGA and buffer layer thickness also change the relationship between interface properties and tunnel seismic response. The ideal interface properties (no-slip and full-slip) will greatly increase or decrease the seismic response of the tunnel. The proposed MRAM and conclusions about the contact interface can provide a reference for the seismic design of the tunnel.

A modified dual time integration technique for aerodynamic quasi-static and dynamic stall hysteresis

Simulation of the aerodynamic stall phenomenon in both quasi-static and dynamic conditions requires expensive computational resources. The computations become even more costly for static stall hysteresis using an unsteady solver with very slow variation of angle of attack at low reduced frequencies. In an explicit time-marching solver that satisfies the low Courant number condition, that is, [Formula: see text], the computational cost for such simulations becomes prohibitive, especially at higher Reynolds numbers due to the presence of thin-stretched cells with large aspect ratio in the boundary layer. In this paper, a segregated solver method such as the Semi-Implicit Method for Pressure-Linked Equations (SIMPLE) is modified as a dual pseudo-time marching method so that the unsteady problem at each time step is reformulated as a steady state problem. The resulting system of equations in the discretized finite volume formulation is then reduced to zero or near-zero residuals using available convergence acceleration methods such as local time stepping, multi-grid acceleration and residual smoothing. The performance and accuracy of the implemented algorithm was tested for three different airfoils at low to moderate Reynolds numbers in both incompressible and compressible flow conditions covering both attached and separated flow regimes. The results obtained are in close agreement with the published experimental and computational results for both quasi-static and dynamic stall and have demonstrated significant savings in computational time.

Achieving maximum speedup in multi-level acceleration for massive coronavirus testing

It is well and widely known that sample pooling could provide an effective and efficient way for fast coronavirus testing among massive asymptomatic individuals. The method of multi-level acceleration for asymptomatic COVID-19 screening has been introduced, and for one and two levels, the optimal group sizes have been obtained. However, there are still multiple challenges. First, it is not clear how to find the optimal group sizes for three or more levels. Second, there is lack of closed-form expressions for the optimal group sizes for two or more levels. Third, it is not clear how to determine the optimal number of levels. And last, it is not known what the maximum achievable speedup is. The motivation of this paper is to address all the above challenges. The optimization of a hierarchical pooling strategy includes its number of levels and the group size of each level. In this paper, based on multi-variable optimization and Taylor approximation, we are able to derive closed-form expressions for the optimal number of levels , the optimal group sizes , ,…, , and the maximum possible speedup of a hierarchical pooling strategy of , where is the fraction of infected people. The above speedup is nearly a linear function of the reciprocal of , in the sense that it is asymptotically greater than any sub-linear function of the reciprocal of for any small . Using the results in this paper, we can quickly and easily predict the performance of an optimal hierarchical pooling strategy. For instance, if the fraction of infected people is 0.0001, an 8-level hierarchical pooling strategy can achieve speedup of nearly 400.