Iteration Time Research Articles

Joint Dynamic Grouping and Gradient Coding for Time-Critical Distributed Machine Learning in Heterogeneous Edge Networks

In edge networks, distributed computing resources have been widely utilized to collaboratively perform a machine learning task by multiple nodes. However, the model training time in heterogeneous edge networks is becoming longer because of excessive computation and delay caused by slow nodes, namely, stragglers. The parameter server even abandons stragglers which fail to return the outcome within a reasonable deadline, called straggler dropout, decreasing the model accuracy. To optimize the computation cost and maintain the model accuracy, we focus on mitigating the heavy computation of stragglers and preventing straggler dropout. Therefore, we propose a novel scheme named dynamic grouping and heterogeneity-aware gradient coding (DGH-GC) to tolerate stragglers by employing dynamic grouping and gradient coding. DGH-GC evenly distributes stragglers in each group and encodes gradients based on their computation capacity to prevent them drop out. However, DGH-GC exacerbates the communication burden by making data duplication to tolerate stragglers. Relying on the scheme, we further propose an algorithm called DGH-(GC)2 to compress transferred gradients in both upstream communication and downstream communication. Experimental evaluations prove that DGH-(GC) outperforms all state-of-the-art methods and DGH-(GC)2 further speeds up the convergence time of the trained model and saves about 26% average iteration time compared to the DGH-(GC).

A Distributed Task Scheduling Method Based on Conflict Prediction for Ad Hoc UAV Swarms

UAV swarms have attracted great attention, and are expected to be used in scenarios, such as search and rescue, that require many urgent jobs to be completed in a minimum time by multiple vehicles. For complex missions with tight constraints, careful assigning tasks is inseparable from the scheduling of these tasks, and multi-task distributed scheduling (MTDS) is required. The Performance Impact (PI) algorithm is an excellent solution for MTDS, but it suffers from the suboptimal solution caused by the heuristics for local task selection, and the deadlock problem that it may fall into an infinite cycle of exchanging the same task. In this paper, we improve the PI algorithm by integrating a new task-removal strategy and a conflict prediction mechanism into the task-removal phase and the task-inclusion phase, respectively. Specifically, the task-removal strategy results in better exploration of the inclusion of more tasks than the original PI by freeing up more space in the local scheduler, improving the suboptimal solution caused by the heuristics for local task selection, as done in PI. In addition, we design a conflict prediction mechanism that simulates adjacent vehicles performing inclusion operations as the criteria for local task inclusion. Therefore, it can reduce the deadlock ratio and iteration times of the MTDS algorithm. Furthermore, by combining the protocol stack with the physical transmission model, an ad-hoc network simulation platform is constructed, which is closer to the real-world network, and serves as the supporting environment for testing the MTDS algorithms. Based on the constructed ad-hoc network simulation platform, we demonstrate the advantage of the proposed algorithm over the original PI algorithm through Monte Carlo simulation of search and rescue tasks. The results show that the proposed algorithm can reduce the average time cost, increase the total allocation number under most random distributions of vehicles-tasks, and significantly reduce the deadlock ratio and the number of iteration rounds.

Large-scale phase retrieval from coded diffraction patterns with electrically tunable lens

General optical detection devices rely on converting photons to electrons (current), and do not allow for direct recording of the phase due to the high oscillation frequency. As so, the missing phase can only be recovered from the intensity measurements. The emerging non-convex phase retrieval algorithm, represented by the Wirtinger Flow (WF) algorithm requires multiple-shot coded diffraction patterns (CDPs) for accurate recovery. To achieve the real-time acquisition for multiple CDPs, this paper proposes a 4f imaging system based on an electrically tunable lens (ETL), which can be used for real-time acquisition of multiple-shot CDPs, and can take the advantages of highspeed, high-resolution, extended depth-of-field, high-sensitivity and low-cost imaging. In this paper, the performance of 4f-ETL based imaging system for phase retrieval with multiple CDPs is compared under different iteration times, different object size, different numbers of masks and different noise levels. Numerical experiments demonstrate the effectiveness and superiority of our proposed ETL-based imaging system, and ETL allows variable-distance focusing of imaging and display systems without mechanical structures, which reduces the mechanical complexity and power consumption, improves acquisition speed.

Global Schrödinger map flows to Kähler manifolds with small data in critical Sobolev spaces: Energy critical case

In this paper and the companion work \[J. Funct. Anal. 281 (2021)], we prove that the Schrödinger map flows from $\mathbb{R}^d$ with $d\ge 2$ to compact Kähler manifolds with small initial data in critical Sobolev spaces are global. The main difficulty compared with the constant sectional curvature case is that the gauged equation now is not self-contained due to the curvature part. Our main idea is to use a novel bootstrap-iteration scheme to reduce the gauged equation to an approximate constant curvature system in finite times of iteration. This paper together with the companion work solves the open problem raised by Tataru.

Optimal error estimates of two‐level iterative finite element methods for the thermally coupled incompressible MHD with different viscosities

In this paper, we develop and design the two‐level finite element iterative methods for the stationary thermally coupled incompressible MHD equations. The considered numerical schemes are based on the classical iterations with one correction. First, under some strong uniqueness conditions, the rough solutions are obtained on the coarse mesh with the mesh size and the Simple, Oseen, and Newton iterations, respectively. Three kinds of corrections are made by solving a linear problem on the fine mesh with mesh size with different viscosities. Finally, under the weak uniqueness condition, the stationary thermally coupled incompressible MHD is solved by the one‐level finite element Oseen iteration on the fine mesh. The uniform stability and convergence of these two‐level iterative methods are analyzed with respect to the mesh sizes and iterative times . Extensive numerical results are presented to demonstrate the established theoretical findings and show the performances of these two‐level iterative schemes with different viscosities.

Research on fault diagnosis of gas steam boilers based on deep neural networks

Gas steam boilers are special equipment subjected to high temperature and pressure, it is particularly important to detect and deal with operation faults in time. Against the 4 types of common fault of gas steam boilers, the fault diagnosis model was established based on deep neural networks (DNN), and the model parameters were optimized and verified by experiments. The results show that the fault diagnosis model performs the best as Tanh of the activation function, 100 of Batch_Size, Adam of Optimizer, 10-5 of learning rate and 0.2 of Dropout since the output accuracy is above 92% and the model begins to converge with iterative times of 12. The accuracy of fault identification of reaches 100% in verification experiments and the model can still effectively recognize the dataset from the early stage which suggests that the fault diagnosis based on deep neural networks has high accuracy and provides a reliable solution for the safety supervision of special equipments.

Finite-time recurrence analysis of chaotic trajectories in Hamiltonian systems.

In this work, we show that a finite-time recurrence analysis of different chaotic trajectories in two-dimensional non-linear Hamiltonian systems provides useful prior knowledge of their dynamical behavior. By defining an ensemble of initial conditions, evolving them until a given maximum iteration time, and computing the recurrence rate of each orbit, it is possible to find particular trajectories that widely differ from the average behavior. We show that orbits with high recurrence rates are the ones that experience stickiness, being dynamically trapped in specific regions of the phase space. We analyze three different non-linear maps and present our numerical observations considering particular features in each of them. We propose the described approach as a method to visually illustrate and characterize regions in phase space with distinct dynamical behaviors.

Flexible-Time Receding Horizon Iterative Learning Control With Application to Marine Hydrokinetic Energy Systems

This brief presents an iterative learning control (ILC) framework for a class of repetitive control (RC) applications characterized by: 1) continuous operation; 2) flexible iteration time; and 3) an economic performance metric. Specifically, the effect of iteration-varying initial conditions, resulting from the continuous nature of the operation, is accounted for through an iteration domain receding horizon formulation. To address the need for flexible iteration times, the time-domain dynamics are transformed into path-domain dynamics characterized by a non-dimensional parameter spanning an iteration-invariant range. The resulting model is used to derive learning filters that minimize a multi-objective economic cost. The proposed methodology is applied to the control a kite-based marine hydrokinetic (MHK) system, which executes high-speed, repetitive flight paths with the objective of maximizing its lap-averaged power output. The proposed approach is validated via simulations of a medium-fidelity nonlinear model of a kite-based MHK system, and the results demonstrate robust and fast convergence of the kite to power-optimal flight patterns.

Short‐term passenger flow prediction for rail transit based on improved particle swarm optimization algorithm

AbstractThe subjectivity of selecting training parameters is an important factor affecting the accuracy of short‐term passenger flow prediction of rail transit by long short‐term memory (LSTM) neural network. In order to improve the prediction accuracy, an improved particle swarm optimization (IPSO) algorithm is proposed to optimize the LSTM. The size of the learning factor of the particle swarm optimization (PSO) algorithm is controlled by dynamic adjustment method to improve the global optimization and convergence ability of the algorithm. The number of hidden layer nodes, learning rate and iteration times of the LSTM are optimized by IPSO. The passenger flow data of Dongjiekou station of Fuzhou Metro Line 1 are selected for verification, and the proposed model is compared with the traditional prediction model. The results show that the LSTM optimized by the improved particle swarm optimization algorithm can effectively predict the short‐term passenger flow of rail transit. Compared with the PSO‐LSTM model, the root mean square error predicted by the IPSO‐LSTM model decreases by 10.87% in the peak period and by 26% in the off‐peak period. The results can provide theoretical and technical support for the optimization of rail transit operation scheme.

UNICADO: multidisciplinary analysis in conceptual aircraft design

In this paper, different iterative methods, so-called architectures, within the multidisciplinary analysis in conceptual aircraft design of the UNICADO software are elaborated, applied and analyzed. Possible execution sequences in the sequential method (Gauss-Seidel architecture) are derived via a graph-based algorithm in combination with expert knowledge. Sensitivities of the design disciplines are analyzed and a permitted residual for stable convergence characteristic for the aircraft design with UNICADO is derived. Prerequisites for the application of a parallel iterative method, the Jacobi architecture are conducted. Runtime and convergence characteristics of the Gauss-Seidel architecture and the Jacobi architecture are evaluated. A damping method is applied to the Jacobi architecture to enhance the convergence characteristics. The Gauss-Seidel and Jacobi architectures are used to design two different aircraft, the CSR and the CSMR, with an iteration accuracy of 2.5e–3. For these use cases studied, the Gauss-Seidel architecture converges more stably and faster than the Jacobi architecture and is, therefore, the more favorable. The aircraft design with the implemented Jacobi architecture oscillates and does not converge. Only with an implemented damping method, convergence is achieved. If the iteration time of the design loop increases, e.g., when using higher fidelity methods for aircraft design, the choice of architecture must be re-evaluated.

Segmentation Algorithm of Magnetic Resonance Imaging Glioma under Fully Convolutional Densely Connected Convolutional Networks.

This work focused on the application value of magnetic resonance imaging (MRI) image segmentation algorithm based on fully convolutional DenseNet neural network (FCDNN) in glioma diagnosis. In this work, based on the fully convolutional DenseNet algorithm, a new MRI image automatic semantic segmentation method cerebral gliomas semantic segmentation network (CGSSNet) was established and was applied to glioma MRI image segmentation by using the BraTS public dataset as research data. Under the same conditions, compare the differences of dice similarity coefficient (DSC), sensitivity, and Hausdroff distance (HD) between this algorithm and other algorithms in MRI image processing. The results showed that the CGSSNet network segmentation algorithm significantly improved the segmentation accuracy of glioma MRI images. In addition, its DSC, sensitivity, and HD values for glioma MRI images were 0.937, 0.811, and 1.201, respectively. Under different iteration times, the DSC, sensitivity, and HD values of the CGSSNet network segmentation algorithm are significantly better than other algorithms. It showed that the CGSSNet model based on the DenseNet can improve the segmentation accuracy of glioma MRI images, and has potential application value in clinical practice.

INTEGRATED STORMWATER ANALYSIS MODEL TO SUPPORT SUSTAINABLE URBAN GREEN SPACE DESIGN

Abstract. Urban Green Space (UGS) has been broadly treated as a valuable and limited resource to handle the challenges brought by high-density urban environment. Green stormwater management has been prompt around the world. To reduce the potential conflicts between stormwater management and other requirements on UGS (especially human needs), the research community and governments encourage the urban planners and designers to integrate stormwater analysis into UGS design. However, the professional term and operation of the traditional hydrological model is a huge challenge to the designers who haven’t touched hydrological knowledge. This study developed a method to simulate and quantify stormwater in UGS by the particle system in the design platform- Rhinoceros +Grasshopper. In this method, we adopted five groups of urban objects such as terrain, building footprint etc. to ease procedures and performance. To overcome the issue the abstracted particle movement cannot reflect the infiltration features of land cover, we introduced categorising locations of the particles and the UGS retaining stormwater hypothesis. To test the feasibility of the method, we tested the three parameters of the integrated model: iteration times, rainfall depth ( selected rainfall event) and particle radius. The comparison tests prove: (1) too small iteration times would lead the particles to stop on the way to the bottom of the terrain, so we set at least 4000 iteration times for simulation; (2) the sensitivity to the selection of rainfall event and particle size is relatively low, the simulation results vary by 1%; (3) too small particle numbers will impact the accuracy of analysis, so we should balance accuracy and work efficiency of work; (4) the stormwater volume estimation based on the particle system is acceptable. The experiments confirm the method can effectively support preliminary UGS design work.

Convolutional neural network based displacement gradients estimation for a full-parameter initial value guess of digital image correlation

The selection of initial value in digital image correlation (DIC) has significant influence on the search efficiency of image subpixel displacement and the algorithmic convergence speed. An accurate and reasonable initial value can reduce the number of iterations of subsequent IC-GN optimization, accelerate the convergence of the results, and avoid the divergence of the algorithm in the iterative process. This paper proposes a full-parameter initial value estimation method based on a regression convolution neural network with multithreaded calculation. The proposed method sequentially uses the integer-pixel estimation based on neighborhood search, the subpixel estimation based on surface fitting and the first-order displacement gradients estimation based on a regressive convolutional neural network to achieve the initial value estimation of inverse compositional Gauss-Newton (IC-GN) iteration. Experimental results show that the iteration times of the proposed method are reduced by about 30% compared with the integer-pixel initial value estimation method. In the process of IC-GN iteration, the computational efficiency of CPU multithreaded calculation is nearly twice higher as that of the single-thread method. It can not only improve the accuracy of the initial value estimation but also has high adaptability, which can adapt to selecting different subset sizes or different speckle patterns. This study provides a reference for the effect of iterative initial value optimization on efficiency and accuracy in DIC.

Construction of Cross-Border e-Commerce Financial Risk Analysis System Based on Support Vector Machine

The rapid changes in the economic situation and the complex market environment make the cross-border e-commerce industry, as an important part of the market economy, face many challenges and risks in the development process. The particularity of its transaction and the imperfect tax mechanism have virtually increased the financial management risk. The research constructs the financial risk analysis model with the help of the support vector machine (SVM) and fuzzy theory. Through algorithm test and empirical research, it is found that the average accuracy of the optimized SVM on the selected data set is more than 90%, and after parameter optimization, the change of model fitness tends to be stable with the increase of iteration times, which greatly improves the search ability of sample data, and the accuracy of financial data classification with high risk is 46.8%. In the empirical research, the model established by fuzzy SVM can effectively eliminate the irrelevant index data, and the prediction accuracy of investment risk and operation risk has reached 80% or more. In the prediction of financing risk and tax risk, its accuracy has been improved by 12.4% compared to that before use.

Research on Music Signal Processing Based on a Blind Source Separation Algorithm

The isolation of mixed music signals is beneficial to the extraction and identification of music signal features and to enhance music signal quality. This paper briefly introduced the mathematical model for separating blind source from mixed music signals and the traditional Independent Component Analysis (ICA) algorithm. The separation algorithm was optimized by the complex neural network. The traditional and optimized ICA algorithms were simulated in MATLAB software. It was found that the time-domain waveform of the signal isolated by the improved ICA-based separation algorithm was closer to the source signal. The similarity coefficient matrix, signal-to-interference ratio, performance index, and iteration time of the improved ICA-based algorithm was 62.3, 0.0011, and 0.87 s, respectively, which were all superior to the traditional ICA algorithm. The novelty of this paper is setting the initial iterative matrix of the ICA algorithm with the complex neural network.

Recovery of a potential coefficient in a pseudoparabolic system from nonlocal observation

We consider a nonlinear inverse problem for identifying an unknown time-dependent potential coefficient in a linear pseudoparabolic equation from nonlocal inversion input, which is quite different from the cases of parabolic equation due to the appearance of the mixed derivatives of third order in the governed equation. Based on the established positive property of the solution to the direct problem, the uniqueness and the Lipschitz conditional stability of this inverse problem are addressed by the principle of contraction mappings for specially chosen weight function in the average observation input data. Then, in terms of the fixed point equation, an iterative algorithm is established to construct the approximate solution to this inverse problem with explicit error estimate, which also leads to the optimal error bounds for specified choice strategy of the iteration times. Some numerical results are presented to validate our proposed reconstruction scheme.

Study and Application of an Elevator Failure Monitoring System Based on the Internet of Things Technology

To prevent the occurrence of elevator safety accidents, in this study, an Internet of things-based elevator failure monitoring system is investigated. First, it introduces the Internet of things technology, preprocesses the relevant data, and extracts the features of the elevator operation data continuously collected by the elevator sensors. The objective of this paper is to study the application of IoT in elevators. Method. The Relief-F algorithm is used to evaluate the potential influencing factors. The results show that, as the batch size increases, the accuracy rate will gradually increase, but after more than 20, the accuracy rate will decrease. When the batch size is 20, the training result is the best. It can be seen that, as the time step increases, the accuracy of the prediction will be significantly improved. When the time step is 24, the prediction accuracy is the highest; after 24, the prediction rate will decline. In the diagram of the influence of learning rate on the model, the blue line indicates that the learning rate is 1.0, the red line is 0.1, and the black line is 0.01. With the increase of the iteration times, the effect is the best when the learning rate is 0.1. This system has high research value and broad application prospects and can make up for the current lack of monitoring in the elevator industry.

An efficient algorithm and complexity result for solving the sum of general affine ratios problem

In this paper, with the unique hypothesis that the denominator is not equal to 0, an efficient outer space rectangle branch-and-bound algorithm is presented to globally solve the sum of general affine ratios problem. By using equivalent transformation and the characteristics of general single ratio function, a linear program relaxation problem is constructed for computing the lower bound of the global minimum value of the original problem. Moreover, to enhance the running speed of the presented algorithm, we design an outer space accelerating technic for deleting the entire outer space rectangle or a part of the entire examined outer space rectangle, in which there contains no the global minimum point of the equivalent problem. Furthermore, through the complexity analysis of the presented algorithm, we estimate its maximum iteration times. In addition, we prove the global convergence of the presented algorithm, report and analyze the numerical computational results for indicating the validity of the presented algorithm. Finally, two practical problems from power transportation and production planning are solved to verify the usefulness of the presented algorithm.

Research on the Integration and Innovation of Ceramic Decorative Pattern Design and Traditional Culture

Ceramics, from the initial practicality to artistry, has a profound impact on social life and its value is unique. Tracing back to the origin of Chinese ceramic art, the application of decorative patterns is legendary. These decorative arts with national and cultural characteristics have distinctive characteristics of the times. Traditional culture is embedded in decorative patterns, which are dominated by the spirit of traditional culture. It is a profound historical accumulation of national traditional culture. Looking at the decorative patterns of ceramic art, these patterns are melting the collision of thoughts, reflecting the ceramic decorative patterns of the light of Chinese civilization for thousands of years, and are also the profound embodiment of the cultural and artistic values of Chinese civilization. This article mainly takes Ganzhou Qilizhen kiln ceramic design as an example to interpret the application of traditional culture in ceramic decorative art. A curve fitting technique based on adaptive genetic algorithm is proposed. The process of curve fitting using adaptive genetic algorithm and curve tools is described in detail, and the comparison results are given. The research shows that the adaptive genetic algorithm can effectively reduce the original iteration times by 30.14% compared with other algorithms. At the same time, this method is applied to the process of pattern design.

Towards Incompressible Laminar Flow Estimation Based on Interpolated Feature Generation and Deep Learning

For industrial design and the improvement of fluid flow simulations, computational fluid dynamics (CFD) solvers offer practical functions and conveniences. However, because iterative simulations demand lengthy computation times and a considerable amount of memory for sophisticated calculations, CFD solvers are not economically viable. Such limitations are overcome by CFD data-driven learning models based on neural networks, which lower the trade-off between accurate simulation performance and model complexity. Deep neural networks (DNNs) or convolutional neural networks (CNNs) are good illustrations of deep learning-based CFD models for fluid flow modeling. However, improving the accuracy of fluid flow reconstruction or estimation in these earlier methods is crucial. Based on interpolated feature data generation and a deep U-Net learning model, this work suggests a rapid laminar flow prediction model for inference of Naiver–Stokes solutions. The simulated dataset consists of 2D obstacles in various positions and orientations, including cylinders, triangles, rectangles, and pentagons. The accuracy of estimating velocities and pressure fields with minimal relative errors can be improved using this cutting-edge technique in training and testing procedures. Tasks involving CFD design and optimization should benefit from the experimental findings.