Massive Computation Research Articles

Optical synaptic devices with ultra-low power consumption for neuromorphic computing

Brain-inspired neuromorphic computing, featured by parallel computing, is considered as one of the most energy-efficient and time-saving architectures for massive data computing. However, photonic synapse, one of the key components, is still suffering high power consumption, potentially limiting its applications in artificial neural system. In this study, we present a BP/CdS heterostructure-based artificial photonic synapse with ultra-low power consumption. The device shows remarkable negative light response with maximum responsivity up to 4.1 × 108 A W−1 at VD = 0.5 V and light power intensity of 0.16 μW cm−2 (1.78 × 108 A W−1 on average), which further enables artificial synaptic applications with average power consumption as low as 4.78 fJ for each training process, representing the lowest among the reported results. Finally, a fully-connected optoelectronic neural network (FONN) is simulated with maximum image recognition accuracy up to 94.1%. This study provides new concept towards the designing of energy-efficient artificial photonic synapse and shows great potential in high-performance neuromorphic vision systems.

Parallelizing the cryo‐EM structure determination in THUNDER using GPU cluster

AbstractElectron cryo‐microscopy (cryo‐EM) is a powerful tool utilized by biologists for understanding the mysteries of life. However, obtaining high‐resolution 3D reconstructions from innumerable noisy images of macromolecules is an extremely complicated task, involving massive image analysis and calculation on a computing cluster. Although extensive efforts have been made for improving the computational efficiency, methods for completely utilizing the computing resources are still challenging for modern cryo‐EM programs. Here, we designed a new computing approach specialized for GPU to optimize and maximize the computing power of a single GPU, multiple GPU, and the GPU cluster, highlighted by a well‐designed cache structure and mixed computing precision of single‐precision and double‐precision. Our approaches achieved remarkable improvement in performance and linear scalability. At an identical cost of the hardware, three‐fold more speed‐up was achieved. The average parallel efficiency can increase up to 84% when multiple GPU configurations are parallelized.

Neural Eikonal solver: Improving accuracy of physics-informed neural networks for solving eikonal equation in case of caustics

The concept of physics-informed neural networks has become a useful tool for solving differential equations due to its flexibility. A few approaches use this concept to solve the eikonal equation that describes the first-arrival traveltimes of waves propagating in smooth heterogeneous velocity models. However, the challenge of the eikonal is exacerbated by the velocity models producing caustics, resulting in instabilities and deterioration of accuracy due to the non-smooth solution behavior. In this paper, we revisit the problem of solving the eikonal equation using neural networks to tackle caustic pathologies. We introduce the novel Neural Eikonal Solver (NES) for solving the isotropic eikonal equation in two formulations: the one-point problem is for a fixed source location; the two-point problem is for an arbitrary source-receiver pair. We present several techniques which provide stability in the case of caustics: improved factorization; non-symmetric loss function based on Hamiltonian; gaussian activation; symmetrization. In our tests, NES showed the relative mean-absolute error of 0.2-0.4% from the second-order factored Fast Marching Method with a similar inference time, and outperformed existing neural-network solvers giving 10-60 times lower errors and 2-30 times faster training. The one-point NES provides the most accurate solution, while the two-point NES gives slightly lower accuracy but an extremely compact representation with all spatial derivatives. It can be useful in many seismic problems: massive computations of traveltimes for millions of source-receiver pairs in Kirchhoff migration; modeling of ray amplitudes using spatial derivatives; traveltime tomography; earthquake localization; ray multipathing analysis.

A Dimension-Reduced Artificial Neural Network Model for the Cell Voltage Consistency Prediction of a Proton Exchange Membrane Fuel Cell Stack

The voltage consistency of hundreds of cells in a proton exchange membrane fuel cell stack significantly influences the stack’s performance and lifetime. Using the physics-based model to estimate the cell voltage consistency is highly challenging due to the massive calculation efforts and the complicated fuel cell designs. In this research, an artificial neural network (ANN) model is developed to efficiently predict the cell voltage distribution and the consistency of a commercial-size fuel cell stack. To balance the computation efficiency and accuracy, a dimension-reduced method is proposed with different output-grouping strategies to optimize the ANN structure based on the experiment test of a 100-cell stack. The model’s training time falls nonlinearly from 16 min to 6 s with the output neuron number decreasing from 100 to 5, while the model can still predict the cell voltage distribution trends. With the proposed model, the stack’s cell voltage distributions could be reproduced with significantly lowered computation time, which is beneficial to evaluate the fuel cell status and optimize the control strategies.

A direct analysis method to Lagrangian global exponential stability for quaternion memristive neural networks with mixed delays

This paper mainly studies the global exponential stability in Lagrange sense (GESLS) of quaternion memristive neural networks (QMNNs) with leakage delays, unbounded distributed delays and time-varying discrete time delays. In the process of research, instead of traditional decomposition into real-valued memristive neural networks (RMNNs) or complex-valued memristive neural networks (CMNNs), we consider the QMNN as a whole, and then give a sufficient condition related to time delays to ensure that the considered QMNN is GESLS. An example is provided to illustrate validity of theoretical results obtained in the end. The method proposed in the present text has two merits: (1) According to the definition of GESLS directly, no Lyapunov–Krasovskii functional (LKF) is required, which avoids massive calculations and solutions of high-dimensional matrix inequalities; (2) It is available not only to QMNNs, but also to RMNNs and CMNNs.

DFSGAN: Introducing editable and representative attributes for few-shot image generation

Training generative adversarial networks (GANs) usually requires large-scale data and massive computation resources. The performance of GANs plummets when given limited data due to the discriminator overfitting, thus providing meaningless feedback to the generator during the adversarial training. Existing few-shot GANs are primarily concerned with transferring knowledge from models that have been pre-trained on large-scale datasets or using data augmentation to expand the training sets. However, previous methods consistently take latent codes sampled from a single distribution as the generator’s input. We contend that more complicated latent codes can provide the generator with more editable attributes. In this paper, we propose DFSGAN for few-shot image generation, which takes dynamic Gaussian mixture (DGM) latent codes as the generator’s input. Our DFSGAN can select the Gaussian components of the latent codes quantitatively. We also design two techniques to strengthen the representative ability of intermediate features of the generating process to improve the fidelity and maintain the content and layout information of the synthesized images. Our DGM and intermediate representation enhancement techniques complement each other and improve synthesis quality. We conduct extensive experiments on 15 few-shot datasets with different resolutions spanning from art paintings to realistic photos. Qualitative and quantitative results demonstrate the superiority and effectiveness of our model.

Efficient graph attentional network for 3D object detection from Frustum-based LiDAR point clouds

LiDAR-based 3D object detection is important for autonomous driving scene perception, but point clouds produced by LiDAR are irregular and unstructured in nature, and cannot be adopted by the conventional Convolutional Neural Networks (CNN). Recently, Graph Convolutional Networks (GCN) has been proved as an ideal way to handle non-Euclidean structure data, as well as for point cloud processing. However, GCN involves massive computation for searching adjacent nodes, and the heavy computational cost limits its applications in processing large-scale LiDAR point cloud in autonomous driving. In this work, we adopt a frustum-based point cloud-image fusion scheme to reduce the amount of LiDAR point clouds, thus making the GCN-based large-scale LiDAR point clouds feature learning feasible. On this basis, we propose an efficient graph attentional network to accomplish the goal of 3D object detection in autonomous driving, which can learn features from raw LiDAR point cloud directly without any conversions. We evaluate the model on the public KITTI benchmark dataset, the 3D detection mAP is 63.72% on KITTI Cars, Pedestrian and Cyclists, and the inference speed achieves 7.9 fps on a single GPU, which is faster than other methods of the same type.

A Decentralized Public Auditing Scheme for Secure Cloud Storage Based on Blockchain

The cloud storage service has brought great convenience to the customer, which can save massive storage and computation resources via outsourcing the data to cloud service provider (CSP). However, the security issues are the biggest challenge such as data integrity. The user can verify the integrity of outsourced data through a remote data auditing solution without retrieving original data from cloud, however, the auditing procedure has heavy computational overhead, which employs third party auditor (TPA) to conduct auditing task on behalf of users. In this paper, we propose a decentralized public auditing scheme for cloud storage based on blockchain, which removes TPA and increases the number of CSP, the auditing task was assigned to multiple CSPs, and the blockchain technology was used to record the audit process. Meanwhile, the structure of e-voting system is utilized to realize the audit result statistics of multiple CSPs via smart contract, which enhanced the credibility and stability of final auditing result. The theoretical analysis and experimental results demonstrate that proposed scheme is secure and efficient.

Secure and Robust Internet of Things with High-Speed Implementation of PRESENT and GIFT Block Ciphers on GPU

With the advent of the Internet of Things (IoT) and cloud computing technologies, vast amounts of data are being created and communicated in IoT networks. Block ciphers are being used to protect these data from malicious attacks. Massive computation overheads introduced by bulk encryption using block ciphers can become a performance bottleneck of the server, requiring high throughput. As the need for high-speed encryption required for such communications has emerged, research is underway to utilize a graphics processor for encryption processing based on the high processing power of the GPU. Applying bit-slicing of lightweight ciphers was not covered in the previous implementation of lightweight ciphers on GPU architecture. In this paper, we implemented PRESENT and GIFT lightweight block ciphers GPU architectures. It minimizes the computation overhead caused by optimizing the algorithm by applying the bit-slicing technique. We performed practical analysis by testing practical use cases. We tested PRESENT-80, PRESENT-128, GIFT-64, and GIFT-128 block ciphers in RTX3060 platforms. The throughput of the exhaustive search are 553.932 Gbps, 529.952 Gbps, 583.859 Gbps, and 214.284 Gbps for PRESENT-80, PRESENT-128, GIFT-64, and GIFT-128, respectively. For the case of data encryption, it achieved 24.264 Gbps, 24.522 Gbps, 85.283 Gbps, and 10.723 Gbps for PRESENT-80, PRESENT-128, GIFT-64, and GIFT-128, respectively. Specifically, the proposed implementation of a PRESENT block cipher is approximately 4× higher performance than the latest work that implements PRESENT block cipher. Lastly, the proposed implementation of a GIFT block cipher on GPU is the first implementation for the server environment.

Operation diagnosis for centrifugal pumps using stator current-based indicators

Motor current signature analysis (MCSA) used in the diagnosis of centrifugal pumps has become a hot topic due to its non-intrusive and cost-effective nature. Yet, traditional current-based diagnosis methods show limitations as massive computation and power frequency interference. Moreover, most of them focus on the fault detection, fewer discuss the applications in operation evaluation. Therefore in this article, the hydro-mechanical-electric coupling effect that exists in the induction motor-centrifugal pump system is analyzed systematically. On this foundation, novel indicators that can reflect the harmonic distortion and noise proportion of stator current signals, and further indicate the hydraulic stability and operation status are proposed. Theoretical and experimental analysis support this paper, and results have demonstrated that our proposed indicators can indicate the operation change, identify the optimal operation region, and reflect the development of cavitation for centrifugal pumps. These indicators have the advantages of easy calculation and identification, therefore can be conveniently compared with accepted standards and enable immediate remedial actions to be implemented to ensure the stable operation.

Configurable Image Rectification and Disparity Refinement for Stereo Vision

Stereo matching is an essential topic in computer vision. However, the high complexity and massive computation restrict its applicability in real-time. This brief presents a pixel-level pipeline coprocessor for configurable binocular stereo rectification, region-optimized semi-global matching (SGM), left-right consistency checking (LRC), and disparity refinement. The contributions are summarized as follows: 1) Instead of storing the entire image of a large number of lines, the buffer size can be significantly reduced by the adjustment range computed according to the pre-calibrated coefficients. 2) An approximate divider with Newton iterations can reach the precision of 10−7 for high accuracy rectification with low latency. 3) A four-layer parallel two-stage pipeline is leveraged to address the computational bottleneck of the SGM, i.e., cost aggregation. 4) The pseudo-dual-port SRAMs and stacks mechanism solve the problem of inverse pixel flow in hole filling with low resource consumption. Finally, the proposed architecture is implemented on an Intel Stratix-IV FPGA and can process 144 frames ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$1024\mathbf {\times }768$ </tex-math></inline-formula> -pixel) per second at a maximum working frequency of 113.13 MHz.

Federated Learning Over Wireless IoT Networks With Optimized Communication and Resources

To leverage massive distributed data and computation resources, machine learning in the network edge is considered to be a promising technique, especially for large-scale model training. Federated learning (FL), as a paradigm of collaborative learning techniques, has obtained increasing research attention with the benefits of communication efficiency and improved data privacy. Due to the lossy communication channels and limited communication resources (e.g., bandwidth and power), it is of interest to investigate fast responding and accurate FL schemes over wireless systems. Hence, we investigate the problem of jointly optimized communication efficiency and resources for FL over wireless Internet of Things (IoT) networks. To reduce complexity, we divide the overall optimization problem into two subproblems, i.e., the client scheduling problem and the resource allocation problem. To reduce the communication costs for FL in wireless IoT networks, a new client scheduling policy is proposed by reusing stale local model parameters. To maximize successful information exchange over networks, a Lagrange multiplier method is first leveraged by decoupling variables, including power variables, bandwidth variables, and transmission indicators. Then, a linear-search-based power and bandwidth allocation method is developed. Given appropriate hyperparameters, we show that the proposed communication-efficient FL (CEFL) framework converges at a strong linear rate. Through extensive experiments, it is revealed that the proposed CEFL framework substantially boosts both the communication efficiency and learning performance of both training loss and test accuracy for FL over wireless IoT networks compared to a basic FL approach with uniform resource allocation.

Event-Driven Computation Offloading in IoT With Edge Computing

Edge computing, which provides computation services at the edge of networks, has become a promising method to meet the massive computation demands of Internet of Things (IoT). To make full use of resources, a computation offloading scheme is needed in edge computing system. In this work, we propose an event-driven computation offloading scheme for the first time. Compared with the existing time-driven schemes, the proposed scheme has a smaller implementation complexity in some scenarios with computation-intensive task computing. In the proposed scheme, the priority of different tasks is jointly considered. To decide the optimal offloading action of the scheme, we formulate the offloading problem as a semi-Markov decision process (SMDP). Then, a model-based method is proposed to derive the optimal offloading policy under fully explored system by addressing the challengs of modeling. On the other hand, considering partially explored system, we propose an online double deep Q-network algorithm, which can deal with the poor scalability of the standard Q-learning algorithm, to derive the optimal offloading policy. In addition, we also introduce some tricks to accelerate the learning procedure. The simulation results show the superior performance of our proposed scheme.

Reliability of arrangement networks in terms of the h-restricted edge connectivity

The h-restricted edge connectivity λh is an important parameter to measure the reliability of massive parallel computing system. The λh of G is the minimum cardinality of a set of edges in G, if any, whose deletion disconnects G, and every remaining component has minimum degree at least h. The arrangement graph is a generalization of the star graph, but its order is more flexible than that of the star graph. This paper investigates the h-restricted edge connectivity of the (n,k)-arrangement graph An,k, and determines that λ1(An,k)=2k(n−k)−2, λ2(An,k)=3k(n−k)−6 with k≤n−2.

Understanding and Applications of Agent-based Model

Agent-based modeling (ABM) is an interdisciplinary approach to understand macroscopic patterns of a large system, based on massive computation of its interacting constituents (i.e., agents). We explain when this approach is especially useful, with providing two game-theoretic examples: The first example is an analytically intractable model system, although the agents’ decision rules are easily programmable, for which ABM is the only feasible methodology. The second example argues that the payoff structure among agents can also be calculated from their microscopic interactions. These examples show that ABM is a powerful tool with a high degree of flexibility, but also that one has to carefully choose the level of complexity in a model because this choice directly affects the computational burden as well as the applicability of the model.

A response reconstruction method based on empirical mode decomposition and modal synthesis method

With the increasing the complexity of civil structures and diversity of the structural components, reconstructing the responses of complex civil engineering structures based on the conventional time-domain reconstruction method is very challenging. Apart from the massive computer storage needed in the calculation process, the enormous calculation times may also hinder the response reconstruction of the civil structures. This paper proposes a novel strategy of the time-domain reconstruction method by combining the modal synthesis method and empirical mode decomposition with intermittency criteria in the time domain. The response reconstruction method is conducted at two levels. The first level is performed on the reduced super-element from the interface DOFs to the internal DOFs of the critical substructure, and then at the element level to extrapolate dynamic responses at critical locations. In each level’s reconstruction, the number of calculation parameters decreases significantly. The response of the whole structure can be quickly obtained at one time, not only decreasing the computation time but also ensuring reconstruction accuracy. The numerical studies and experimental validation are both conducted on a simply supported beam and a complex cooling tower for verification, and the effects from some external parameters, including measurement noise, the order of substructural normal modes, the number of substructures and the number of sensors, are considered. The results illustrate that the proposed reconstruction method based on the modal synthesis method improves the calculation effectiveness considerably and ensures accuracy.

Bulirsh–Stoer algorithm in the planar restricted three-body problem

The applicability of the Bulirsh–Stoer algorithm for solving the planar restricted three-body problem is investigated. Variations in the value of the Jacobi integral are considered as the main parameter. Massive calculations were carried out with a small step in the parameter characterizing the ratio of the masses of a planet and star (the mass parameter). It is shown that violations of the Jacobi integral occur inside the planetary chaotic region. This fact can be used to determine chaotic region boundaries, as well as the boundaries of a stable structure which coorbital with the planet. The average dependences of the value of the Jacobi integral on the mass parameter, which determine the boundaries of the chaotic zone and the coorbital ring, are derived. Estimates were obtained for the maximum relative change in the Jacobi integral for different values of the accretion radius. It is shown that the value of the accretion radius corresponding to the average radius of known exoplanets does not cause significant changes in the value of the Jacobi integral. The dependences of the clearing time of the chaotic zone for different accretion radii are also given with and without taking into account the coorbital structure.

Users' experience matter: Delay sensitivity-aware computation offloading in mobile edge computing

As a promising computing paradigm, Mobile Edge Computing (MEC) provides communication and computing capability at the edge of the network to address the concerns of massive computation requirements, constrained battery capacity and limited bandwidth of the Internet of Things (IoT) systems. Most existing works on mobile edge task ignores the delay sensitivities, which may lead to the degraded utility of computation offloading and dissatisfied users. In this paper, we study the delay sensitivity-aware computation offloading by jointly considering both user's tolerance towards delay of task execution and the network status under computation and communication constraints. Specifically, we use a specific multi-user and multi-server MEC system to define the latency sensitivity of task offloading based on the analysis of delay distribution of task categories. Then, we propose a scoring mechanism to evaluate the sensitivity-dependent utility of task execution and devise a Centralized Iterative Redirection Offloading (CIRO) algorithm to collect all information in the MEC system. By starting with an initial offloading strategy, the CIRO algorithm enables IoT devices to cooperate and iteratively redirect task offloading decisions to optimize the offloading strategy until it converges. Extensive simulation results show that our method can significantly improve the utility of computation offloading in MEC systems and has lower time complexity than existing algorithms.

Efficient Massive Computing for Deformable Volume Data Using Revised Parallel Resampling

In this paper, we propose an improved parallel resampling technique. Parallel resampling is a deformable object generation method based on volume data applied to medical simulations. Existing parallel resampling is not suitable for massive computing, because the number of samplings is high and floating-point precision problems may occur. This study addresses these problems to obtain improved user latency when performing medical simulations. Specifically, instead of interpolating values after volume sampling, the efficiency is improved by performing volume sampling after coordinate interpolation. Next, the floating-point error in the calculation of the sampling position is described, and the advantage of barycentric interpolation using a reference point is discussed. The experimental results showed a significant improvement over the existing method. Volume data comprising more than 600 images used in clinical practice were deformed and rendered at interactive speed. In an Internet of Everything environment, medical imaging systems are an important application, and simulation image generation is also valuable in the overall system. Through the proposed method, the performance of the whole system can be improved.

Configuration design of hybrid constellations with synchronous orbital precession based on rapid coverage analysis

This paper mainly deals with the configuration design of hybrid constellations for Earth discontinuous coverage. Firstly, a general design model of the hybrid constellation with synchronous orbital precession and multiple repeating ground tracks is developed. Then, to cope with the massive coverage calculation during the constellation design process, an analytical method for rapid coverage analysis of hybrid constellations is proposed. Finally, numerical examples are carried out, and several hybrid constellation design cases are proposed. Results indicate that the proposed coverage analysis method is demonstrated to be of high precision, and the designed hybrid constellations can achieve superior coverage performance compared with conventional Walker constellations. This research can be conducive to hybrid constellations design and Earth discontinuous coverage analysis.