Improvement In Execution Time Research Articles

CMAN: Cascaded Multi-scale Spatial Channel Attention-guided Network for large 3D deformable registration of liver CT images

Deformable image registration is an essential component of medical image analysis and plays an irreplaceable role in clinical practice. In recent years, deep learning-based registration methods have demonstrated significant improvements in convenience, robustness and execution time compared to traditional algorithms. However, registering images with large displacements, such as those of the liver organ, remains underexplored and challenging. In this study, we present a novel convolutional neural network (CNN)-based unsupervised learning registration method, Cascaded Multi-scale Spatial-Channel Attention-guided Network (CMAN), which addresses the challenge of large deformation fields using a double coarse-to-fine registration approach. The main contributions of CMAN include: (i) local coarse-to-fine registration in the base network, which generates the displacement field for each resolution and progressively propagates these local deformations as auxiliary information for the final deformation field; (ii) global coarse-to-fine registration, which stacks multiple base networks for sequential warping, thereby incorporating richer multi-layer contextual details into the final deformation field; (iii) integration of the spatial-channel attention module in the decoder stage, which better highlights important features and improves the quality of feature maps. The proposed network was trained using two public datasets and evaluated on another public dataset as well as a private dataset across several experimental scenarios. We compared CMAN with four state-of-the-art CNN-based registration methods and two well-known traditional algorithms. The results show that the proposed double coarse-to-fine registration strategy outperforms other methods in most registration evaluation metrics. In conclusion, CMAN can effectively handle the large-deformation registration problem and show potential for application in clinical practice. The source code is made publicly available at https://github.com/LocPham263/CMAN.git.

Migration-aware slot-based memory request scheduler to guarantee QoS in DRAM-PCM hybrid memories

Emerging hybrid memory technologies composed of non-volatile memories (NVM) and DRAMs exhibit significant access speeds and capacity improvement. High application performance is feasible by dynamic migration (or relocation) of pages (data) between the memory types. While NVM is used for its density during memory allocation, moving write-intensive pages from NVM to DRAM helps to improve the execution and response time of applications. Existing techniques propose solutions to dynamically identify the pages that need to be moved immediately or in regular intervals. Such an immediate or interval-based rigid migration regime may hamper the service of the regular memory requests, which in turn affects the memory service rate.To alleviate the impact on service rate and improve the quality-of-service (QoS) of the device, this paper proposes a scheduling method to identify the instant at which to migrate the eligible page. Along with regular memory requests, these eligible migration candidate pages are given reserved time slots by taking into account the current memory request rate. Our proposed methods aim to optimize the migration overheads by avoiding unnecessary migrations and at the same time guaranteeing future accesses to the migrated pages. This results in improved execution time and memory response time for the applications.

RING 4.0: faster residue interaction networks with novel interaction types across over 35,000 different chemical structures.

Residue interaction networks (RINs) are a valuable approach for representing contacts in protein structures. RINs have been widely used in various research areas, including the analysis of mutation effects, domain-domain communication, catalytic activity, and molecular dynamics simulations. The RING server is a powerful tool to calculate non-covalent molecular interactions based on geometrical parameters, providing high-quality and reliable results. Here, we introduce RING 4.0, which includes significant enhancements for identifying both covalent and non-covalent bonds in protein structures. It now encompasses seven different interaction types, with the addition of π-hydrogen, halogen bonds and metal ion coordination sites. The definitions of all available bond types have also been refined and RING can now process the complete PDB chemical component dictionary (over 35000 different molecules) which provides atom names and covalent connectivity information for all known ligands. Optimization of the software has improved execution time by an order of magnitude. The RING web server has been redesigned to provide a more engaging and interactive user experience, incorporating new visualization tools. Users can now visualize all types of interactions simultaneously in the structure viewer and network component. The web server, including extensive help and tutorials, is available from URL: https://ring.biocomputingup.it/.

VerSA: Versatile Systolic Array Architecture for Sparse and Dense Matrix Multiplications

A key part of modern deep neural network (DNN) applications is matrix multiplication. As DNN applications are becoming more diverse, there is a need for both dense and sparse matrix multiplications to be accelerated by hardware. However, most hardware accelerators are designed to accelerate either dense or sparse matrix multiplication. In this paper, we propose VerSA, a versatile systolic array architecture for both dense and sparse matrix multiplications. VerSA employs intermediate paths and SRAM buffers between the rows of the systolic array (SA), thereby enabling an early termination in sparse matrix multiplication with a negligible performance overhead when running dense matrix multiplication. When running sparse matrix multiplication, 256 × 256 VerSA brings performance (i.e., an inverse of execution time) improvement and energy saving by 1.21×–1.60× and 7.5–30.2%, respectively, when compared to the conventional SA. When running dense matrix multiplication, VerSA results in only a 0.52% performance overhead compared to the conventional SA.

Sharing Queries with Nonequivalent User-defined Aggregate Functions

This article presents Sharing User-Defined Aggregate Function (SUDAF), a declarative framework that allows users to write User-defined Aggregate Functions (UDAFs) as mathematical expressions and use them in Structured Query Language statements. SUDAF rewrites partial aggregates of UDAFs using built-in aggregate functions and supports efficient dynamic caching and reusing of partial aggregates. Our experiments show that rewriting UDAFs using built-in functions can significantly speed up queries with UDAFs, and the proposed sharing approach can yield up to two orders of magnitude improvement in query execution time. The article studies also an extension of SUDAF to support sharing partial results between arbitrary queries with UDAFs. We show a connection with the problem of query rewriting using views and introduce a new class of rewritings, called SUDAF rewritings, which enables to use views that have aggregate functions different from the ones used in the input query. We investigate the underlying rewriting-checking and rewriting-existing problem. Our main technical result is a reduction of these problems to, respectively, rewriting-checking and rewriting-existing of the so-called aggregate candidates , a class of rewritings that has been deeply investigated in the literature.

ExTern: Boosting RISC-V core performance using ternary encoding

This paper presents an effective μ-architectural design method, called ExTern, to enhance the performance of a RISC-V processor experiencing computation bottlenecks. ExTern involves integrating Canonical Signed Digit (CSD) representation, a ternary number system enabling carry/borrow-free addition/subtraction in constant time O(1), into the RISC-V processor, particularly into the execution stage. Furthermore, it adopts an extended six-stage pipeline architecture to maximize employed encoding benefits, leading to more improvement in overall execution time and throughput. Despite the presence of optimized circuits, such as fast carry chain (CARRY4) modules for binary encoding on FPGA, the customized processor applying ExTern, RISC-VT, showcases remarkable improvement in computation performance. Experimental results demonstrate a 34.3% (12.2%) improvement in working frequency leading to a lower 31% (11.5%) execution time and a 32% (12%) increase in throughput compared to a State-of-the-Art open-source five(six)-stage RISC-V processor.

Reliable and Accurate Implicit Neural Representation of Multiple Swept Volumes with Application to Safe Human–Robot Interaction

In automated production using collaborative robots in a manufacturing cell, a crucial aspect is to avoid collisions to ensure the safety of workers and robots in human–robot interaction. One approach for detecting collisions is using the swept volume (SV) to identify safe protective space for operation. We learn an accurate and reliable signed distance function (SDF) network from raw point clouds of a pre-computed SV to represent a class of linear joint motion trajectories. The network requires only a set of parameters and constant execution time, thus reducing the computational time and memory of collision checking due to the complexity of explicit geometry during task execution. The distance of collision danger foresaw by the learned SDF is exploited to reduce the frequency of collision detection calls in the dynamic environment and reduce the computational cost further. We assess the relative merits of the implicit neural representation of multiple SVs in terms of F1-score, error distance from the surface of the truth geometry, and 3D visualization by comparing favorably with a binary voxel network for learning a single SV with similar inference time. All the predicted errors of the geometry lie within a distance of 4 voxels from the surface of the truth geometry, and most reconstruction errors are within 3 voxels. A simulation of pick-and-place task execution in the human–robot interaction scenarios by leveraging the learned SDF as an efficient continuous collision detector is performed. The improvement in execution time and collision detection number is validated in the simulation.

A novel deep learning technique for medical image analysis using improved optimizer.

Application of Convolutional neural network in spectrum of Medical image analysis are providing benchmark outputs which converges the interest of many researchers to explore it in depth. Latest preprocessing technique Real ESRGAN (Enhanced super resolution generative adversarial network) and GFPGAN (Generative facial prior GAN) are proving their efficacy in providing high resolution dataset. Objective: Optimizer plays a vital role in upgrading the functioning of CNN model. Different optimizers like Gradient descent, Stochastic Gradient descent, Adagrad, Adadelta and Adam etc. are used for classification and segmentation of Medical image but they suffer from slow processing due to their large memory requirement. Stochastic Gradient descent suffers from high variance and is computationally expensive. Dead neuron problem also proves to detrimental to the performance of most of the optimizers. A new optimization technique Gradient Centralization is providing the unparalleled result in terms of generalization and execution time. Method: Our paper explores the next factor which is the employment of new optimization technique, Gradient centralization (GC) to our integrated framework (Model with advanced preprocessing technique). Result and conclusion: Integrated Framework of Real ESRGAN and GFPGAN with Gradient centralization provides an optimal solution for deep learning models in terms of Execution time and Loss factor improvement.

Gaussian process regression and conditional Karhunen-Loève models for data assimilation in inverse problems

We present a model inversion algorithm, CKLEMAP, for data assimilation and parameter estimation in partial differential equation models of physical systems with spatially heterogeneous parameter fields. These fields are approximated using low-dimensional conditional Karhunen-Loève expansions (CKLEs), which are constructed using Gaussian process regression (GPR) models of these fields trained on the parameters' measurements. We then assimilate measurements of the state of the system and compute the maximum a posteriori (MAP) estimate of the CKLE coefficients by solving a nonlinear least-squares problem. When solving this optimization problem, we efficiently compute the Jacobian of the vector objective by exploiting the sparsity structure of the linear system of equations associated with the forward solution of the physics problem.The CKLEMAP method provides better scalability compared to the standard MAP method. In the MAP method, the number of unknowns to be estimated is equal to the number of elements in the numerical forward model. On the other hand, in CKLEMAP, the number of unknowns (CKLE coefficients) is controlled by the smoothness of the parameter field and the number of measurements, and is generally much smaller than the number of discretization nodes, which leads to a significant reduction of computational cost with respect to the standard MAP method. To show this advantage in scalability, we apply CKLEMAP to estimate the transmissivity field in a two-dimensional steady-state subsurface flow model of the Hanford Site by assimilating synthetic measurements of transmissivity and hydraulic head. We find that the execution time of CKLEMAP scales nearly linearly as N1.33, where N is the number of discretization nodes, while the execution time of standard MAP scales as N2.91. The CKLEMAP method improved execution time without sacrificing accuracy when compared to the standard MAP method.

An Optimal Random Forest Model for Enhancing Decision-Making in Improving Students’ Performance via Educational Data Mining

The need for measures to enhance students’ performance, especially in Science and Math as a panacea for development in this 4th industrial revolution has become a global call. The large stream and heterogeneous nature of educational data make it difficult to use traditional statistical methods for data analysis to support decisions. However, Educational data mining has been effective and efficient in addressing this issue but much focus of researchers has been on only students' academic records to determine performance. This study sought to propose a Random Forest model to predict and enhance students’ performance. The study adopted a 5-stage mining process to mine psycho-socio-economic demographics educational data with Mutual information, Chi-squared test, Featurewiz, RandomizedSearch CV, and GridSearch CV to optimize the model’s performance. The study’s outcome revealed the key factors affecting students’ performance and that the model was enhanced by a 10.4% increment in precision and f1-score and 9.1% recall value, 7.1secs (62%) improvement in execution time and 78.7% improvement in Root Mean Square Error. This outcome remains a contribution to guiding decision-making in the educational setting and a basis for further studies on model optimization.

Exploring Classification of Topological Priors With Machine Learning for Feature Extraction.

In many scientific endeavors, increasingly abstract representations of data allow for new interpretive methodologies and conceptualization of phenomena. For example, moving from raw imaged pixels to segmented and reconstructed objects allows researchers new insights and means to direct their studies toward relevant areas. Thus, the development of new and improved methods for segmentation remains an active area of research. With advances in machine learning and neural networks, scientists have been focused on employing deep neural networks such as U-Net to obtain pixel-level segmentations, namely, defining associations between pixels and corresponding/referent objects and gathering those objects afterward. Topological analysis, such as the use of the Morse-Smale complex to encode regions of uniform gradient flow behavior, offers an alternative approach: first, create geometric priors, and then apply machine learning to classify. This approach is empirically motivated since phenomena of interest often appear as subsets of topological priors in many applications. Using topological elements not only reduces the learning space but also introduces the ability to use learnable geometries and connectivity to aid the classification of the segmentation target. In this article, we describe an approach to creating learnable topological elements, explore the application of ML techniques to classification tasks in a number of areas, and demonstrate this approach as a viable alternative to pixel-level classification, with similar accuracy, improved execution time, and requiring marginal training data.

Improve Multi-Object Detection and Tracking for an Automated Traffic Surveillance System

Multi-Object Detection and Tracking (MODT) are essential in many application fields. Still, many enhancements in the speed of detection and tracking were required to overcome the challenges during implementation. This paper presents a new algorithm system for (MODT) to improve the execution time to be robust in real-time applications. A background subtraction detection algorithm with a Kalman filter was used to track and predict the object position and speed parameters. To improve the processing time, its needs to reduce some frames in a way that does not affect the detection accuracy too much and instead use the prediction and the estimated value obtained based on the Kalman filter for the tracked object. This work uses a single video camera to show how effectively to compute and detect multiple objects concurrently; it is applied for daytime preprocessing in an automated traffic surveillance system. Preliminary testing findings show that the suggested algorithm for this vehicle monitoring system is feasible and effective. It illustrates that using the suggested algorithm with a single video camera can simultaneously watch, detect, and track several vehicles and improve execution time. Simulation results on the built system demonstrate that the proposed system reduced the execution time to approximately 41.5% compared to the standard background subtraction algorithm. Results indicate the proposed algorithm has an approximate error for the position and speed of detected and tracked objects compared with the standard background subtraction algorithm. Index Terms— multi-object detection, tracking, background subtraction, Kalman filter, morphological.

너비 우선 탐색 기반 비재귀 트리 탐색 알고리즘의 효율적 구현과 성능 비교

The objective of this study is to present an efficient tree traversal algorithm in terms of execution time and implementation complexity. Trees, defined as acyclic connected graphs, are widely used in various fields of computer science. Unlike traditional recursive methods or depth first search implementations, this research proposes an implementation based on breadth first search. The algorithm improves execution time using a non-recursive approach and reduces implementation complexity by using only one-dimensional arrays and loops. Additionally, the algorithm has been optimized through a preprocessing phase. To evaluate the performance of the proposed algorithm compared to existing ones, trees with random structures were generated and their execution times were measured. The results demonstrated that the proposed algorithm exhibits efficient performance.

GPU-accelerated full-field modelling of highly dispersive ultrafast optical parametric oscillators

We demonstrate GPU-accelerated modelling of ultrafast optical parametric oscillators (OPOs) via the χ(2) nonlinear envelope equation with 1265× improvement in execution time compared with a CPU-based approach. Incorporating an adaptive step-size algorithm and absorbing boundary conditions, our model is capable of simulating OPOs containing long (>10 mm) nonlinear crystals or significant intracavity dispersion with outputs generated in less than 1 minute, allowing the investigation of systems that were previously computationally prohibitive to explore. We implement real-world parameters such as optical coatings, material absorption, and non-ideal poling domains within quasi-phase matched nonlinear crystals, producing excellent agreement with the spectral tuning behaviour and average power from a previously reported prism-based OPO. Our digital twinning approach provides a low-cost iterative development platform for ultrafast OPOs.

A Harris Hawk Optimisation system for energy and resource efficient virtual machine placement in cloud data centers.

Virtualisation is a major technology in cloud computing for optimising the cloud data centre's power usage. In the current scenario, most of the services are migrated to the cloud, putting more load on the cloud data centres. As a result, the data center's size expands resulting in increased energy usage. To address this problem, a resource allocation optimisation method that is both efficient and effective is necessary. The optimal utilisation of cloud infrastructure and optimisation algorithms plays a vital role. The cloud resources rely on the allocation policy of the virtual machine on cloud resources. A virtual machine placement technique, based on the Harris Hawk Optimisation (HHO) model for the cloud data centre is presented in this paper. The proposed HHO model aims to find the best place for virtual machines on suitable hosts with the least load and power consumption. PlanetLab's real-time workload traces are used for performance evaluation with existing PSO (Particle Swarm Optimisation) and PABFD (Best Fit Decreasing). The performance evaluation of the proposed method is done using power consumption, SLA, CPU utilisation, RAM utilisation, Execution time (ms) and the number of VM migrations. The performance evaluation is done using two simulation scenarios with scaling workload in scenario 1 and increasing resources for the virtual machine to study the performance in underloaded and overloaded conditions. Experimental results show that the proposed HHO algorithm improved execution time(ms) by 4%, had a 27% reduction in power consumption, a 16% reduction in SLA violation and an increase in resource utilisation by 17%. The HHO algorithm is also effective in handling dynamic and uncertain environments, making it suitable for real-world cloud infrastructures.

DBSCAN inspired task scheduling algorithm for cloud infrastructure

Cloud computing in today's computing environment plays a vital role, by providing efficient and scalable computation based on pay per use model. To make computing more reliable and efficient, it must be efficient, and high resources utilized. To improve resource utilization and efficiency in cloud, task scheduling and resource allocation plays a critical role. Many researchers have proposed algorithms to maximize the throughput and resource utilization taking into consideration heterogeneous cloud environments. This work proposes an algorithm using DBSCAN (Density-based spatial clustering) for task scheduling to achieve high efficiency. The proposed DBScan-based task scheduling algorithm aims to improve user task quality of service and improve performance in terms of execution time, average start time and finish time. The experiment result shows proposed model outperforms existing ACO and PSO with 13% improvement in execution time, 49% improvement in average start time and average finish time. The experimental results are compared with existing ACO and PSO algorithms for task scheduling.

An Optimized Hardware Implementation of Modular Multiplication of Binary Ring LWE

Providing end-to-end security is vital for most networks. Emerging quantum computers make it necessary to design secure crypto-systems against quantum attacks. Binary Ring Learning With Error (Ring-Bin LWE) is a Lattice-based cryptography that is hard to solve by quantum computers. Also, this algorithm does not have costly operations in terms of area, making Ring-Bin LWE a suitable algorithm for resource-constraint devices. This work presents a lightweight hardware implementation of Ring-Bin LWE. In the proposed design, a new multiplication method and design for Ring-Bin LWE is introduced which results in latency reduction by a factor of two. Using column-based multiplication, our design processes two consecutive coefficients in each cycle. The architecture is designed based on the proposed multiplication and contains one specific register bank with two sub-bank registers. The design is implemented on the FPGA platforms. The implementation results show an impressive improvement in execution time and Area-Time metrics over previous similar works.

Communication-hiding pipelined BiCGSafe methods for solving large linear systems

Recently, a new variant of the BiCGStab method, known as the pipelined BiCGStab, has been proposed. This method can achieve a higher degree of scalability and speed-up rates through a mechanism in which the communication phase for the computation of the inner product can be overlapped with the computation of the matrix-vector product. Meanwhile, several generalized iteration methods with better convergence behavior than BiCGStab exist, such as ssBiCGSafe, BiCGSafe, and GPBi-CG. Among these methods, ssBiCGSafe, which requires a single phase of computing inner products per iteration, is best suited for high-performance computing systems. As described herein, inspired by the success of the pipelined BiCGStab method, we propose pipelined variations of the ssBiCGSafe method in which only one phase of inner product computation per iteration is required and this phase of inner product computation can be overlapped with the matrix-vector computation. Through numerical experimentation, we demonstrate that the proposed methods engender improvements in convergence behavior and execution time compared to the pipelined BiCGStab and ssBiCGSafe methods.

Human Activity Recognition Using Cascaded Dual Attention CNN and Bi-Directional GRU Framework.

Vision-based human activity recognition (HAR) has emerged as one of the essential research areas in video analytics. Over the last decade, numerous advanced deep learning algorithms have been introduced to recognize complex human actions from video streams. These deep learning algorithms have shown impressive performance for the video analytics task. However, these newly introduced methods either exclusively focus on model performance or the effectiveness of these models in terms of computational efficiency, resulting in a biased trade-off between robustness and computational efficiency in their proposed methods to deal with challenging HAR problem. To enhance both the accuracy and computational efficiency, this paper presents a computationally efficient yet generic spatial-temporal cascaded framework that exploits the deep discriminative spatial and temporal features for HAR. For efficient representation of human actions, we propose an efficient dual attentional convolutional neural network (DA-CNN) architecture that leverages a unified channel-spatial attention mechanism to extract human-centric salient features in video frames. The dual channel-spatial attention layers together with the convolutional layers learn to be more selective in the spatial receptive fields having objects within the feature maps. The extracted discriminative salient features are then forwarded to a stacked bi-directional gated recurrent unit (Bi-GRU) for long-term temporal modeling and recognition of human actions using both forward and backward pass gradient learning. Extensive experiments are conducted on three publicly available human action datasets, where the obtained results verify the effectiveness of our proposed framework (DA-CNN+Bi-GRU) over the state-of-the-art methods in terms of model accuracy and inference runtime across each dataset. Experimental results show that the DA-CNN+Bi-GRU framework attains an improvement in execution time up to 167× in terms of frames per second as compared to most of the contemporary action-recognition methods.

A Cost-Effective Multi-Verse Optimization Algorithm for Efficient Power Generation in a Microgrid

Renewable energy sources (RESs) are a great source of power generation for microgrids with expeditious urbanization and increase in demand in the energy sector. One of the significant challenges in deploying RESs with microgrids is efficient energy management. Optimizing the power allocation among various available generation units to serve the load is the best way to achieve efficient energy management. This paper proposes a cost-effective multi-verse optimizer algorithm (CMVO) to solve this optimization problem. CMVO focuses on the optimal sharing of generated power in a microgrid between different available sources to reduce the generation cost. The proposed algorithm is analyzed for two different scale microgrids (IEEE 37-node test system and IEEE 141-node test system) using IEEE test feeder standards to assess its performance. The results show that CMVO outperforms multi-verse optimizer (MVO), particle swarm optimization (PSO), artificial hummingbird algorithm (AHA), and genetic algorithm (GA). The simulation results emphasize the cost reduction and execution time improvement in both IEEE test systems compared with other meta-heuristic algorithms.