Global Computation Research Articles

In-Vitro Flow Validation of Third-Generation Ventricular Assist Devices: Feasibility and Challenges.

Computational fluid dynamics (CFD) is a powerful tool for the in-silico evaluation of rotodynamic blood pumps (RBPs). Corresponding validation, however, is typically restricted to easily accessible, global flow quantities. This study showcased the HeartMate 3 (HM3) to identify feasibility and challenges of enhanced in-vitro validation in third-generation RBPs. To enable high-precision acquisition of impeller torques and grant access for optical flow measurements, the HM3 testbench geometry was geometrically modified. These modifications were reproduced in silico , and global flow computations validated along 15 operating conditions. The globally validated flow in the testbench geometry was compared with CFD-simulated flows in the original geometry to assess the impact of the necessary modifications on global and local hydraulic properties. Global hydraulic properties in the testbench geometry were successfully validated (pressure head: r = 0.999, root mean square error [RMSE] = 2.92 mmHg; torque: r = 0.996, RMSE = 0.134 mNm). In-silico comparison with the original geometry demonstrated good agreement ( r > 0.999, relative errors < 11.97%) of global hydraulic properties. Local hydraulic properties (errors up to 81.78%) and hemocopatibility predictions (deviations up to 21.03%), however, were substantially affected by the geometric modifications. Transferability of local flow measures derived on advanced in-vitro testbenches toward original pump designs is challenged by significant local effects associated with the necessary geometrical modifications.

Enhancing the blockchain interoperability through federated learning with directed acyclic graph

AbstractThe use of federated learning to achieve blockchain interoperability has become a hot topic in research, because it enables data exchange without revealing any private information. However, the previous work, such as ScaleSFL (Asia‐CCS, 2022), that has implemented federated learning for blockchain interoperability, the throughput of the framework still cannot support the practical applications. Therefore, a federated learning framework based on Directed Acyclic Graph (DAG) is proposed which utilizes sharding mechanism to enhance the blockchain interoperability. By constructing a weighted context graph based on data correlation, reasonable sharding of the dataset is achieved, thereby improving the efficiency of blockchain interoperability. The experimental results show that the federated framework reduces global computation in federated learning by 30% compared with other schemes, while increasing blockchain throughput by nearly 40%.

A Fully Integrated Pipeline for Visual Carotid Morphology Analysis

AbstractAnalyzing stenoses of the internal carotids – local constrictions of the artery – is a critical clinical task in cardiovascular disease treatment and prevention. For this purpose, we propose a self‐contained pipeline for the visual analysis of carotid artery geometries. The only inputs are computed tomography angiography (CTA) scans, which are already recorded in clinical routine. We show how integrated model extraction and visualization can help to efficiently detect stenoses and we provide means for automatic, highly accurate stenosis degree computation. We directly connect multiple sophisticated processing stages, including a neural prediction network for lumen and plaque segmentation and automatic global diameter computation. We enable interactive and retrospective user control over the processing stages. Our aims are to increase user trust by making the underlying data validatable on the fly, to decrease adoption costs by minimizing external dependencies, and to optimize scalability by streamlining the data processing. We use interactive visualizations for data inspection and adaption to guide the user through the processing stages. The framework was developed and evaluated in close collaboration with radiologists and neurologists. It has been used to extract and analyze over 100 carotid bifurcation geometries and is built with a modular architecture, available as an extendable open‐source platform.

Expected local topology of random complex submanifolds

Let n ≥ 2 n\geq 2 and r ∈ { 1 , ⋯ , n − 1 } r\in \{1, \cdots , n-1\} be integers, M M be a compact smooth Kähler manifold of complex dimension n n , E E be a holomorphic vector bundle with complex rank r r and equipped with a Hermitian metric h E h_E , and L L be an ample holomorphic line bundle over M M equipped with a metric h h with positive curvature form. For any d ∈ N d\in \mathbb N large enough, we equip the space of holomorphic sections H 0 ( M , E ⊗ L d ) H^0(M,E\otimes L^d) with the natural Gaussian measure associated to h E h_E , h h and its curvature form. Let U ⊂ M U\subset M be an open subset with smooth boundary. We prove that the average of the ( n − r ) (n-r) -th Betti number of the vanishing locus in U U of a random section s s of H 0 ( M , E ⊗ L d ) H^0(M,E\otimes L^d) is asymptotic to ( n − 1 r − 1 ) d n ∫ U c 1 ( L ) n {n-1 \choose r-1} d^n\int _U c_1(L)^n for large d d . On the other hand, the average of the other Betti numbers is o ( d n ) o(d^n) . The first asymptotic recovers the classical deterministic global algebraic computation. Moreover, such a discrepancy in the order of growth of these averages is new and contrasts with all known other smooth Gaussian models, in particular the real algebraic one. We prove a similar result for the affine complex Bargmann-Fock model.

A neural network based global traveltime function (GlobeNN)

Global traveltime modeling is an essential component of modern seismological studies with a whole gamut of applications ranging from earthquake source localization to seismic velocity inversion. Emerging acquisition technologies like distributed acoustic sensing (DAS) promise a new era of seismological discovery by allowing a high-density of seismic observations. Conventional traveltime computation algorithms are unable to handle virtually millions of receivers made available by DAS arrays. Therefore, we develop GlobeNN—a neural network based traveltime function that can provide seismic traveltimes obtained from the cached realistic 3-D Earth model. We train a neural network to estimate the traveltime between any two points in the global mantle Earth model by imposing the validity of the eikonal equation through the loss function. The traveltime gradients in the loss function are computed efficiently using automatic differentiation, while the P-wave velocity is obtained from the vertically polarized P-wave velocity of the GLAD-M25 model. The network is trained using a random selection of source and receiver pairs from within the computational domain. Once trained, the neural network produces traveltimes rapidly at the global scale through a single evaluation of the network. As a byproduct of the training process, we obtain a neural network that learns the underlying velocity model and, therefore, can be used as an efficient storage mechanism for the huge 3-D Earth velocity model. These exciting features make our proposed neural network based global traveltime computation method an indispensable tool for the next generation of seismological advances.

Resolving the Horizontal Direction of Internal Tide Generation: Global Application for the M2 Tide’s First Mode

Abstract Breaking internal tides contribute substantially to small-scale turbulent mixing in the ocean interior and hence to maintaining the large-scale overturning circulation. How much internal tide energy is available for ocean mixing can be estimated by using semianalytical methods based on linear theory. Until recently, a method resolving the horizontal direction of the internal waves generated by conversion of the barotropic tide was lacking. We here present the first global application of such a method to the first vertical mode of the principal lunar semidiurnal internal tide. We also show that the effect of supercritical slopes on the modally decomposed internal tides is different than previously suggested. To deal with this the continental shelf and the shelf slope are masked in the global computation. The global energy conversion obtained agrees roughly with the previous results by Falahat et al. if the mask is applied to their result, which decreases their energy conversion by half. Thus, around half of the energy conversion obtained by their linear calculations occurs at continental slopes and shelves, where linear theory tends to break down. The barotropic-to-baroclinic energy flux at subcritical slopes away from the continental margins is shown to vary substantially with direction depending on the shape and orientation of topographic obstacles and the direction of the local tidal currents. Taking this additional information into account in tidal mixing parameterizations could have important ramifications for vertical mixing and water mass properties in global numerical simulations.

Field-Circuit Coupling With Element-by-Element FEM

An element-by-element finite element method (EBE-FEM) for coupling fields and circuits is presented. EBE-FEM is a parallel algorithm in which large-scale global computations are decomposed into mutually independent local computations without assembling the global coefficient matrix. The electromagnetic correlation matrix is proposed to establish the relationship between each element and the whole domain in the magnetic field equation and the relationship between element and branch in circuit equation, which is composed by node transition matrix and branch transition matrix. The proposed method has been validated and applied to analyze the magnetic field and currents of a transformer.

A Dynamic Grid Index for CkNN Queries on Large-Scale Road Networks with Moving Objects

As the Internet of Things devices are deployed on a large scale, location-based services are being increasingly utilized. Among these services, kNN (k-nearest neighbor) queries based on road network constraints have gained importance. This study focuses on the CkNN (continuous k-nearest neighbor) queries for non-uniformly distributed moving objects with large-scale dynamic road network constraints, where CkNN objects are continuously and periodically queried based on their motion evolution. The present CkNN high-concurrency query under the constraints of a super-large road network faces problems, such as high computational cost and low query efficiency. The aim of this study is to ensure high concurrency nearest neighbor query requests while shortening the query response time and reducing global computation costs. To address this issue, we propose the DVTG-Index (Dynamic V-Tree Double-Layer Grid Index), which intelligently adjusts the index granularity by continuously merging and splitting subgraphs as the objects move, thereby filtering unnecessary vertices. Based on DVTG-Index, we further propose the DVTG-CkNN algorithm to calculate the initial kNN query and utilize the existing results to speed up the CkNN query. Finally, extensive experiments on real road networks confirm the superior performance of our proposed method, which has significant practical applications in large-scale dynamic road network constraints with non-uniformly distributed moving objects.

Block-partitioned Rayleigh–Ritz method for efficient eigenpair reanalysis of large-scale finite element models

Abstract In this manuscript, we propose a new effective method for eigenpair reanalysis of large-scale finite element (FE) models. Our method utilizes the matrix block-partitioning algorithm in the Rayleigh–Ritz approach and expresses the Ritz basis matrix using thousands of block matrices of very small size. To avoid significant computational costs from the projection procedure, we derive a new formulation that uses tiny block computations instead of global matrix computations. Additionally, we present an algorithm that recognizes which blocks are changed in the modified FE model to achieve computational cost savings when computing new eigenpairs. Through selective updating for the recognized blocks, we can effectively construct the new Ritz basis matrix and the new reduced mass and stiffness matrices corresponding to the modified FE model. To demonstrate the performance of our proposed method, we solve several practical engineering problems and compare the results with those of the combined approximation method, the most well-known eigenpair reanalysis method, and ARPACK, an eigenvalue solver embedded in many numerical programs.

DHT: Dynamic Vision Transformer Using Hybrid Window Attention for Industrial Defect Images Classification

Industrial defect detection is gaining importance in the control of industrial product quality. Highly accurate and efficient defect detection with complex and variable industrial defect types is therefore an interesting but challenging problem. Vision transformers have been highly successful in a variety of computer vision tasks, due to their ability to capture global information in images. Nevertheless, simply capturing global information is problematic. On the one hand, because they are incapable of inductive bias as Convolutional Neural Network (CNN), transformers will have difficulty focusing on local features of defects in industrial defect image inspection tasks. On the other hand, using global computation leads to excessive memory and computational cost. To mitigate these issues, we propose a new vision transformer architecture which contains Hybrid Window Attention (HWA) and Dynamic Token Normalization (DTN). HWA, which combines pooling attention and window attention, makes the computational complexity reduced to improve efficiency. DTN enables transformers to focus on both the global information and the local features of defects, thus providing improved accuracy of industrial surface defect detection. Extensive experiments demonstrate that our Dynamic Vision Transformer (DHT) achieves 96.8% and 98.5% classification accuracy on the NEU dataset and the DAGM dataset, respectively, with a low computational complexity.

Bridging the Gap Between Cumbersome and Light Detectors via Layer-Calibration and Task-Disentangle Distillation in Remote Sensing Imagery

With urgent application requirements such as satellite in-orbit processing and unmanned aerial vehicle tracking, knowledge distillation (KD) following the teacher-student teaching mechanism has shown great potential to obtain lightweight detectors. However, compact students have limited accuracy due to the interference of large scale variations and blurred boundaries in remote sensing objects. Specifically, previous methods mostly force teacher-student responses from the layer of same depth and scale to align. Stereotyped manual inter-layer associations may cause discriminative features of multi-scale objects to be incorrectly bundled. Further, the regression branch follows the identical distillation paradigm as the classification branch, resulting in ambiguous object bounding box deviations. To solve the above two issues, we propose an effective KD framework called layer-calibration and task-disentangle distillation (LTD). First, cross-layer calibration distillation (CCD) structure is innovatively proposed. It adaptively binds a student layer with several related target layers, rather than a fixed layer in the teacher model. Appropriate and clear knowledge of large and small objects is transmitted. Since the CCD structure requires explicit global inner product computation between multiple layers, local implicit calibration (LIC) module is further proposed to reduce distilled convergence difficulty. Second, task-aware spatial disentangle distillation (TASD) structure is devised to transfer task-decoupled semantics and localization knowledge in a divide-and-conquer manner, alleviating objects localization imprecision. Experiments demonstrate that our LTD achieves state-of-the-art performance on several datasets, and is a plug-and-play approach to most detectors. The code will be available soon.

Exploiting More Information in Sparse Point Cloud for 3D Single Object Tracking

3D single object tracking is a key task in 3D computer vision. However, the sparsity of point clouds makes it difficult to compute the similarity and locate the object, posing big challenges to the 3D tracker. Previous works tried to solve the problem and improved the tracking performance in some common scenarios, but they usually failed in some extreme sparse scenarios, such as for tracking objects at long distances or partially occluded. To address the above problems, in this letter, we propose a sparse-to-dense and transformer-based framework for 3D single object tracking. First, we transform the 3D sparse points into 3D pillars and then compress them into 2D bird's eye view (BEV) features to have a dense representation. Then, we propose an attention-based encoder to achieve global similarity computation between template and search branches, which could alleviate the influence of sparsity. Meanwhile, the encoder applies the attention on multi-scale features to compensate for the lack of information caused by the sparsity of point cloud and the single scale of features. Finally, we use set-prediction to track the object through a two-stage decoder which also utilizes attention. Extensive experiments show that our method achieves very promising results on the KITTI and NuScenes datasets.

High-resolution multi-domain space–time isogeometric analysis of car and tire aerodynamics with road contact and tire deformation and rotation

We are presenting high-resolution space–time (ST) isogeometric analysis of car and tire aerodynamics with near-actual tire geometry, road contact, and tire deformation and rotation. The focus in the high-resolution computation is on the tire aerodynamics. The high resolution is not only in space but also in time. The influence of the aerodynamics of the car body comes, in the framework of the Multidomain Method (MDM), from the global computation with near-actual car body and tire geometries, carried out earlier with a reasonable mesh resolution. The high-resolution local computation, carried out for the left set of tires, takes place in a nested MDM sequence over three subdomains. The first subdomain contains the front tire. The second subdomain, with the inflow velocity from the first subdomain, is for the front-tire wake flow. The third subdomain, with the inflow velocity from the second subdomain, contains the rear tire. All other boundary conditions for the three subdomains are extracted from the global computation. The full computational framework is made of the ST Variational Multiscale (ST-VMS) method, ST Slip Interface (ST-SI) and ST Topology Change (ST-TC) methods, ST Isogeometric Analysis (ST-IGA), integrated combinations of these ST methods, element-based mesh relaxation (EBMR), methods for calculating the stabilization parameters and related element lengths targeting IGA discretization, Complex-Geometry IGA Mesh Generation (CGIMG) method, MDM, and the “ST-C” data compression. Except for the last three, these methods were used also in the global computation, and they are playing the same role in the local computation. The ST-TC, for example, as in the global computation, is making the ST moving-mesh computation possible even with contact between the tire and the road, thus enabling high-resolution flow representation near the tire. The CGIMG is making the IGA mesh generation for the complex geometries less arduous. The MDM is reducing the computational cost by focusing the high-resolution locally to where it is needed and also by breaking the local computation into its consecutive portions. The ST-C data compression is making the storage of the data from the global computation less burdensome. The car and tire aerodynamics computation we present shows the effectiveness of the high-resolution computational analysis framework we have built for this class of problems.

Exhaustive Exploitation of Local Seeding Algorithms for Community Detection in a Unified Manner

Community detection is an essential task in network analysis and is challenging due to the rapid growth of network scales. Recently, discovering communities from the local perspective of some specified nodes called seeds, rather than requiring the global information of the entire network, has become an alternative approach to addressing this challenge. Some seeding algorithms have been proposed in the literature for finding seeds, but many of them require an excessive amount of effort because of the global information or intensive computation involved. In our study, we formally summarize a unified framework for local seeding by considering only the local information of each node. In particular, both popular local seeding algorithms and new ones are instantiated from this unified framework by adopting different centrality metrics. We categorize these local seeding algorithms into three classes and compare them experimentally on a number of networks. The experiments demonstrate that the degree-based algorithms usually select the fewest seeds, while the denseness-based algorithms, except the one with node mass as the centrality metric, select the most seeds; using the conductance of the egonet as the centrality metric performs best in discovering communities with good quality; the core-based algorithms perform best overall considering all the evaluation metrics; and among the core-based algorithms, the one with the Jaccard index works best. The experimental results also reveal that all the seeding algorithms perform poorly in large networks, which indicates that discovering communities in large networks is still an open problem that urgently needs to be addressed.

Eccentricity queries and beyond using hub labels

Hub labeling schemes are popular methods for computing distances on road networks and other large complex networks, often answering to a query within a few microseconds for graphs with millions of edges. In this work, we study their algorithmic applications beyond distance queries. Indeed, several implementations of hub labels were reported to have good practical performances, both in terms of pre-processing time and maximum label size. There are also a few relevant graph classes for applications for which we know how to compute hub labelings with sublinear labels in quasi linear time. These positive results raise the question of what can be computed efficiently given a graph and a hub labeling of small maximum label size as input. We focus on eccentricity queries and distance-sum queries, for several versions of these problems on directed weighted graphs, that is in part motivated by their importance in facility location problems. On the negative side, we show conditional lower bounds for these above problems on unweighted undirected sparse graphs, via standard constructions from “Fine-grained” complexity. Specifically, under the Strong Exponential-Time Hypothesis (SETH), answering to these above queries requires Ω(|V|2−o(1)) pre-processing time or Ω(|V|1−o(1)) query time, even if we are given as input a hub labeling of maximum label size in ω(log⁡|V|). However, things take a different turn when the hub labels have a sublogarithmic size. Indeed, for every ε>0 there exists a δε>0 such that the following hold, being given a hub labeling of maximum label size ≤k: after pre-processing the labels in total O(2δε⋅k⋅|V|1+ε) time, we can compute both the eccentricity and the distance-sum of any vertex in O(2δε⋅k⋅|V|ε) time. Our data structure is a novel application of the orthogonal range query framework of Cabello and Knauer (2009) [17]. It can also be applied to the fast global computation of some topological indices. Finally, as a by-product of our approach, on any fixed class of unweighted graphs with bounded expansion, we can decide whether the diameter of an n-vertex graph in the class is at most k in fε(k)⋅n1+ε time for any ε>0, for some “explicit” function fε that depends on ε and the graph class considered. This result is further motivated by the empirical evidence that many classes of complex networks have bounded expansion (Demaine et al., 2019 [24]), and that some of these networks, such as the online social networks, have a relatively small diameter.

Factors Affecting the Adoption of IoT-Based Smart Campus: An Investigation Using Analytical Hierarchical Process (AHP)

The advancement of technology is making university campuses smarter every single day. Despite the benefits of these advanced technologies, the literature concerning the adoption of smart campuses is significantly lacking increased knowledge to provide effective smart campus solutions. This study aims to prioritize the adoption factors of an IoT-based smart campus. The study applied an analytical hierarchical process (AHP) on 25 factors drawn from the literature. The factors were classified into technology specific factors (TSF), organizational specific factors (OSF), environmental specific factors (ESF), and end-user specific factors (USF). Based on the results obtained, the most significant contributing factors were government support, privacy concerns, social influence, facilitating conditions, and service collaboration, whereas the least significant contributing factors were enjoyment, availability, reliability, mobility, and compatibility. Moreover, based on the global ranking computation, 12 factors from the OSF, ESF, and USF categories appeared to be more significant than TSF. The findings of this study could help university administrators, manufacturers, and policy-makers to understand the critical factors of smart campuses in order to improve the adoption and utilization of these solutions effectively.

A heuristic whale optimization algorithm with niching strategy for global multi-dimensional engineering optimization

Whale optimization algorithm (WOA) has received increasing attention in engineering optimization owing to its high computation efficiency, whereas, it has exhibited the drawback of premature convergence in solving multi-dimensional engineering global optimization problems. In this research, a niching hybrid heuristic whale optimization algorithm (NHWOA) is proposed to enhance convergence speed and search coverage in solving global optimization problems. In the algorithm, the niching technique is introduced to promote the diversity of population and restrain premature convergence in search of a global best solution. A heuristic adjustment to the parameters of the hybrid WOA algorithm is made to promote the exploration potential of search agents in the evolution. A designed perturbation to all the search agents’ positions is executed to avoid their falling into a local optimum. Optimization to the CEC2014 benchmark functions as validation cases are conducted along with comparisons to both conventional intelligent optimization algorithms and other state-of-the-art modified WOAs. Results indicate the effectiveness and superiority of NHWOA in solving the problems. Five practical engineering problems for global optimization with multiply variables are introduced to validate the performance of the presented algorithm with good performance results in the global computations.

Twin delayed deep deterministic policy gradient-based intelligent computation offloading for IoT

In view of the randomness distribution of multiple users in the dynamic large-scale Internet of Things (IoT) scenario, comprehensively formulating available resources for fog nodes in the area and achieving computation services at low cost have become great challenges. As a result, this paper studies an efficient and intelligent computation offloading mechanism with resource allocation. Specifically, an optimization problem is formulated to minimize the total energy consumption of all tasks under the joint optimization of computation offloading decisions, bandwidth resources and transmission power. Meanwhile, a Twin Delayed Deep Deterministic Policy Gradient-based Intelligent Computation Offloading (TD3PG-ICO) algorithm is proposed to solve this optimization problem. By combining the concept of the actor critic algorithm, the proposed algorithm designs two independent critic networks that can avoid the subjective prediction of a single critic network and better guide the policy network to generate the global optimal computation offloading policy. Additionally, this algorithm introduces a continuous variable discretization operation to select the target offloading node with random probability. The available resources of the target node are dynamically allocated to improve the model decision-making effect. Finally, the simulation results show that this proposed algorithm has faster convergence speed and good robustness. It can always approach the greedy algorithm with respect to the lowest total energy consumption. Furthermore, compared with full local and Deep Q-learning Network (DQN)-based computation offloading schemes, the total energy consumption can be reduced by an average of 15.53% and 6.41%, respectively.

Sparse non-negative matrix factorization for uncertain data clustering

We consider the problem of clustering a set of uncertain data, where each data consists of a point-set indicating its possible locations. The objective is to identify the representative for each uncertain data and group them into k clusters so as to minimize the total clustering cost. Different from other models, our model does not assume that there is a probability distribution for each uncertain data. Thus, all possible locations need to be considered to determine the representative. Existing methods for this problem are either impractical or have difficulty to handle large-scale datasets due to their pairwise-distance based global search strategy and expensive optimization computation. In this paper, we propose a novel sparse Non-negative Matrix Factorization (NMF) method which measures the similarity of uncertain data by their most commonly shared features. A divide-and-conquer approach is adopted to remarkably improve the efficiency. A novel diagonal l0-constraint and its l1 relaxation are proposed to overcome the challenge of determining the representatives. We give a detailed analysis to show the correctness of our method, and provide an effective initialization and peeling strategy to enhance the ability of processing large-scale datasets. Experimental results on some benchmark datasets confirm the effectiveness of our method.

A New Proof of Gaffney’s Inequality for Differential Forms on Manifolds-with-Boundary: The Variational Approach à La Kozono-Yanagisawa

Let (\({\cal M}\), g0) be a compact Riemannian manifold-with-boundary. We present a new proof of the classical Gaffney inequality for differential forms in boundary value spaces over \({\cal M}\), via a variational approach à la Kozono-Yanagisawa [Lr-variational inequality for vector fields and the Helmholtz-Weyl decomposition in bounded domains, Indiana Univ. Math. J. 58 (2009), 1853–1920], combined with global computations based on the Bochner technique.