Execution Time Research Articles

Modelos de predicción para detectar pacientes con enfermedades cardiacas empleando redes neuronales y las bibliotecas Pytorch y TensorFlow

Neural Networks are used to generate prediction models aimed at determining whether an individual has heart conditions. The PyTorch and TensorFlow libraries are applied to a dataset of patients related to heart diseases, containing 17 predictor variables. The purpose of this work is to compare the results obtained using the aforementioned libraries with Neural Networks, analyzing the behavior of loss functions and the outcomes from the confusion matrix when creating the prediction model. DownSampling and UpSampling techniques are employed to address the imbalance in the dataset, which consists of a total of 319,795 patients, of whom only 27,373 have heart disease. It was found that for this dataset, the best results with PyTorch are achieved in models of 100 epochs and above, with execution times of only a few seconds, while TensorFlow shows good results starting from models with 10 epochs, though its execution time is considerably longer. An analysis is conducted on the difference in computation time between PyTorch and TensorFlow.

Adaptive reverse task offloading in edge computing for AI processes

Nowadays, we witness the proliferation of edge IoT devices, ranging from smart cameras to autonomous vehicles, with increasing computing capabilities, used to implement AI-based services in users’ proximity, right at the edge. As these services are often computationally demanding, the popular paradigm of offloading their tasks to nearby cloud servers has gained much traction and been studied extensively. In this work, we propose a new paradigm that departs from the above typical edge computing offloading idea. Namely, we argue that it is possible to leverage these end nodes to assist larger nodes (e.g., cloudlets) in executing AI tasks. Indeed, as more and more end nodes are deployed, they create an abundance of idle computing capacity, which, when aggregated and exploited in a systematic fashion, can be proved beneficial. We introduce the idea of reverse offloading and study a scenario where a powerful node splits an AI task into a group of subtasks and assigns them to a set of nearby edge IoT nodes. The goal of each node is to minimize the overall execution time, which is constrained by the slowest subtask, while adhering to predetermined energy consumption and AI performance constraints. This is a challenging MINLP (Mixed Integer Non-Linear Problem) optimization problem that we tackle with a novel approach through our newly introduced EAI-ARO (Edge AI-Adaptive Reverse Offloading) algorithm. Furthermore, a demonstration of the efficacy of our reverse offloading proposal using an edge computing testbed and a representative AI service is performed. The findings suggest that our method optimizes the system’s performance significantly when compared with a greedy and a baseline task offloading algorithm.

An efficient test case selection framework for multitenant SaaS applications using RTS4MTS

The augmentation of multitenant Software as a Service (SaaS) has significantly contributed to the success of cloud computing. To stay competitive, multitenant SaaS applications are frequently updated, necessitating regression tests. Regression testing ensures modifications do not alter existing functionalities or introduce new bugs. Rerunning all test cases is resource intensive. To minimize efforts, it's crucial to select a relevant subset of test cases. Thus, a novel framework, Regression Test Selection for Multitenant SaaS (RTS4MTS), is proposed for selective regression testing in multitenant SaaS applications. This framework uses the Functional Coverage Module (FCM) and Dependency Analysis Module (DAM) to select relevant test cases. It classifies existing test cases into two groups: those affected by modifications (need to test) and those unaffected (no need to test). Applying RTS4MTS to the multitenant SaaS MtBookTrip application demonstrated that the framework reduces the test suite size by 68% and lowers test execution time by 70%, thereby saving resources and testing costs.

Securing the internet of things frontier: a deep learning ensemble for cyber-attack detection in smart environments

<span lang="EN-US">This study presents a novel and innovative approach using deep learning (DL) ensemble technique to improve the security of internet of things (IoT) by identifying intricate cyber-attacks. By utilising advanced DL models like deep neural network (DNN) and long short-term memory (LSTM), our approach significantly enhances the accuracy of categorization compared to basic models. The initial binary classifier achieved an accuracy of 85.2%, while the multi-class classifier achieved an accuracy of 79.7%. Both classifiers continually enhanced, achieving accuracies of 99.34% and 98.26%, respectively, after 100 epochs. Real-time scenario evaluations showed that the average execution time per sample record was 0.9439 ms, confirming its efficiency. The DL ensemble exhibited improved performance in comparison to traditional models, indicating its potential for wider implementation in IoT security. The study not only emphasises significant improvements in accuracy, but also emphasises the method’s ability to perform well across many evaluation measures. This study presents a thorough and pragmatic method for identifying cyber-attacks in IoT settings. The stacked ensemble technique outperforms earlier models and fulfils real-time processing requirements, offering substantial advancements in IoT security. These findings enhance both the theoretical comprehension and practical application, establishing a novel benchmark for protecting intelligent IoT systems.</span>

Memory Optimization using Dynamic Programming: A Comprehensive

Abstract: Dynamic Programming (DP) is one of the main techniques applied in problem solving by breaking a complex problem into its small, easier parts. It focuses on how these parallel methods are applied in current parallel computing systems, especially shared memory systems using OpenMP and distributed memory systems using MPI. We analyzed and compared execution times, scalability, and communication costs in which these parallel DP methods perform well. Shared memory systems are easier to implement for small to medium problems because they have low communication costs. Distributed memory systems are better suited for large problems though distributed memory systems have higher communication costs. This gives one an idea of the strengths and weaknesses associated with different parallel DP methods. It also enables the choice of the most relevant methods according to the features of the computing problem

Comparative analysis of machine learning models for predicting healthcare traffic: Insights for optimized emergency response

Efficient management of healthcare traffic is crucial for ensuring timely access to medical services, particularly in emergency situations where delays can have severe consequences. This study presents a comparative analysis of three widely used machine learning models—Linear Regression, Decision Trees, and Random Forests—aimed at predicting healthcare-related traffic volumes. A large dataset from a metropolitan traffic system was used to train and evaluate the models based on key performance indicators, including Mean Squared Error (MSE), R² Score, and computational efficiency. The results reveal that the Random Forest model offers the best performance, achieving higher predictive accuracy and faster execution times compared to the other models. These findings provide valuable insights into the use of machine learning for optimizing healthcare traffic management, potentially enhancing response times and improving patient outcomes.

Enhancing Transcriptomic Analysis by Influencing De Novo Assembly Using Parallel Computing

The efficient and accurate assembly of genomic data is a computationally intensive process that demands significant computational resources. Traditional sequential approaches often struggle to handle genomic data sets increasing volume and complexity, leading to prolonged execution times and suboptimal results. The study aims to leverage parallel computing capabilities by employing the ABySS and Velvet Assembler tools on the MD2 Pineapple dataset hosted on the Quanta server. By systematically evaluating the performance of these tools across varying thread counts, the study seeks to identify optimal configurations that can enhance the efficiency and accuracy of the de novo assembly process, ultimately enabling more rapid and precise genomic analysis. The study found that for the ABySS assembler, an 8-core and 8-thread configuration exhibited the shortest execution time and greatest speedup, while an 8-core and 12-thread setup produced similar outcomes, demonstrating ABySS's flexibility to adjust to various thread configurations. Velvet assembler demonstrated exceptional performance by utilizing 8 cores and 16 threads for the velvetg command, and 8 cores and 8 threads for the velveth command. Significantly, this study provided implications for advancing genomic analysis methodologies by providing valuable guidance on optimizing the efficiency and accuracy of de novo assembly processes through careful selection of parallelization configurations, paving the way for future studies and applications in genetic data analysis.

Analysis of Network Security Measures for Protecting Organizational Data and Implement Digital Ideas

Objectives: The goal of this study is to proposed analysis of Efficacious Intellectual Framework for Host based IDS improves the quality of the Association Rule Mining. Methods: For these 15 examiners were appropriated. To create great reactions, a solitary survey is utilized to quantify numerous hypothetical builds of the review. The poll is separated into three fundamental segments, where, the main area manages general inquiries. In the subsequent segment, the respondents are mentioned to pick one explicit elective that presents the situation with given secuity practice in their organization. The Herberos convention was utilized in this review. Results: Brute-Force method is used to reduce the total running time of the framework when the proposed framework with suggested IRG method is used for unknown rule generation and single keyword pattern matching. The proposed framework consumes less memory and less execution time. Conclusions: This approach is taken advantage of to increase the expectation precision. In this commitment, a creative Bright Characterization Structure is wanted to uplift the exactness of the grouping in the IDS. The Proposed PFSM technique prompted seize the unimportant component in the IDS dataset for the characterization of the organization. From the result achieved, it is validated that the proposed approach achieved its uniqueness over the normal element determination method executed in the Interruption Discovery Framework. Additionally, it likewise progresses the forecast exactness and reduces the mistake rates. The minimization of blunder rates is the result of incredible order precision.

Exploring K-Means Clustering Efficiency: Accuracy and Computational Time across Multiple Datasets

Exploring K-Means Clustering Efficiency: Accuracy and Computational Time across Multiple Datasets

Moving Beyond Oxford Nanopore Standard Procedures: New Insights from Water and Multiple Fish Microbiomes

Oxford Nanopore Technology (ONT) allows for the rapid profiling of aquaculture microbiomes. However, not all the experimental and downstream methodological possibilities have been benchmarked. Here, we aimed to offer novel insights into the use of different library preparation methods (standard-RAP and native barcoding-LIG), primers (V3–V4, V1–V3, and V1–V9), and basecalling models (fast-FAST, high-HAC, and super-accuracy-SUP) implemented in ONT to elucidate the microbiota associated with the aquatic environment and farmed fish, including faeces, skin, and intestinal mucus. Microbial DNA from water and faeces samples could be amplified regardless of the library–primer strategy, but only with LIG and V1–V3/V1–V9 primers in the case of skin and intestine mucus. Low taxonomic assignment levels were favoured by the use of full-length V1–V9 primers, though in silico hybridisation revealed a lower number of potential matching sequences in the SILVA database, especially evident with the increase in Actinobacteriota in real datasets. SUP execution allowed for a higher median Phred quality (24) than FAST (11) and HAC (17), but its execution time (6–8 h) was higher in comparison to the other models (0.6–7 h). Altogether, we optimised the use of ONT for water- and fish-related microbial analyses, validating, for the first time, the use of the LIG strategy. We consider that LIG–V1–V9-HAC is the optimal time/cost-effective option to amplify the microbial DNA from environmental samples. However, the use of V1–V3 could help to maximise the dataset microbiome diversity, representing an alternative when long amplicon sequences become compromised by microbial DNA quality and/or high host DNA loads interfere with the PCR amplification/sequencing procedures, especially in the case of gut mucus.

Mesh Smoothing of 3D Prosthesis and Orthosis Models

Introduction. Formation of a contact surface in prosthetics and orthotics is crucial for the restoration of human musculoskeletal functions. This paper considers specific features of methods currently used for denoising of a 3D model surface obtained by optical scanning. An algorithm for manufacturing an individual prosthetic and orthopedic product is developed. The current literature reports no similar methods, which may be explained by the widespread use of gypsum technology for the manufacture of prostheses and orthoses.Aim. Research and development of digital filtration methods for 3D meshes obtained by optical scanning for further modeling of individual prosthesis and orthosis modules.Materials and methods. It is proposed to optimize the pre-processing stage of 3D scanned models by applying denoising and smoothing processes. In total, 50 optical 3D scans were selected for testing via the following denoising algorithms: bilateral filtering, vertex-based anisotropic smoothing, mean and median filtering applied to face normals.Results. The study was conducted using 3D scans of lower limb stumps and Chenault brace orthoses and corsets provided by the Institute of Prosthetics and Orthotics, Albrecht Federal Scientific and Educational Centre of Medical and Social Expertise and Rehabilitation. А method for denoising and smoothing of the 3D model surfaces for the manufacture of prosthetic and orthopedic products is proposed. The SNR metric difference SNR – δ-SNR (with averaging by scans and SNR values) and average execution time were calculated. The bilateral filtration method with δ-SNR = 11.3362 dB and a runtime of 8.8900 s showed the highest efficiency.Conclusion. The proposed methods for the pre-processing stage of 3D optical scans showed high efficiency in the formation of 3D models of prosthesis and orthosis modules. The results obtained can be used for automating the process of manufacturing various prosthetic and orthopedic products, which is particularly relevant in the context of the modern geopolitical situation.

Ex-vivo evaluation of a percutaneous thread technique to palmar/plantar annular desmotomy in horses

Palmar/plantar annular desmotomy (PAD) is a surgical procedure to mitigate pain, lameness, and other clinical manifestations of annular ligament syndrome. This study aimed to describe a minimally invasive technique for PAD and evaluate potential damage to adjacent structures in 15 cadaveric limbs from sound horses (forelimb = 6 and hindlimb = 9). A 6 USP polyglycolic acid thread was positioned around the palmar/plantar annular ligament (PAL) on both sides of the limbs through two incisions (proximal and distal to PAL) through the digital sheath and subcutaneous tissue. Kelly forceps, Gerlach or Obwegeser needles were used to thread placement. A back-and-forth motion allowed the thread to cut the ligament until it exited through the proximal skin incision, and the time was registered. The degree of PAL section and tissue injuries were evaluated. The section of the ligament was complete in all limbs. However, 30% (10/30) of the procedures produced iatrogenic injuries on adjacent structures regardless of the instrument used (p > 0.1). There was a higher frequency of lesions in the neurovascular bundle (NVB) on the medial side regardless of the limb (p = 0.068). The odds of injury in the NVB were correlated to side (lateral or medial (p = 0.042) and limb (hind or forelimb) (p = 0.065) by logistic regression. The Gerlach needle had a shorter execution time than the Kelly forceps (p = 0.003). The percutaneous approach to PAD in specimens was practical, quick, and easy to perform. However, this posed a potential chance of iatrogenic trauma to the surrounding tissue. Although they are all reasonable options, the Obwegeser needle was more efficient for applying the thread than the others.

Enhancing keratoconus detection with transformer technology and multi-source integration

Keratoconus is a progressive eye disease characterized by the thinning and conical distortion of the cornea, leading to visual impairment. Early and accurate detection is essential for effective management and treatment. Traditional diagnostic methods, relying primarily on corneal topography, often fail to detect early-stage keratoconus due to their subjective nature and limited scope. In this research, we present a novel multi-source detection approach utilizing transformer technology to predict keratoconus progression more accurately. By integrating and analyzing diverse data sources, including corneal topography, aberrometry, pachymetry, and biomechanical properties, our method captures subtle changes indicative of disease progression. Transformer networks, known for their capability to model complex dependencies in data, are employed to handle the multimodal datasets effectively. Experimental results demonstrate that our approach significantly outperforms existing methods, such as SVM-based, Random Forests-based, and CNN-based models, in terms of accuracy, precision, recall, and F-score. Moreover, the proposed system exhibits lower execution times, highlighting its efficiency in clinical settings. This innovative methodology holds the potential to revolutionize keratoconus management by enabling earlier and more precise interventions, ultimately enhancing patient outcomes and contributing significantly to both the medical and machine learning communities.

Sparse Neighbor Joining: rapid phylogenetic inference using a sparse distance matrix.

Phylogenetic reconstruction is a fundamental problem in computational biology. The Neighbor Joining (NJ) algorithm offers an efficient distance-based solution to this problem, which often serves as the foundation for more advanced statistical methods. Despite prior efforts to enhance the speed of NJ, the computation of the n 2 entries of the distance matrix, where n is the number of phylogenetic tree leaves, continues to pose a limitation in scaling NJ to larger datasets. In this work, we propose a new algorithm which does not require computing a dense distance matrix. Instead, it dynamically determines a sparse set of at most O(n log n) distance matrix entries to be computed in its basic version, and up to O(n log 2n) entries in an enhanced version. We show by experiments that this approach reduces the execution time of NJ for large datasets, with a trade-off in accuracy. Sparse Neighbor Joining is implemented in Python and freely available at https://github.com/kurtsemih/SNJ. Supplementary data are available at Bioinformatics online.

UAV-Mounted RIS-Aided Multi-Target Localization System: An Efficient Sparse-Reconstruction-Based Approach

Unmanned Aerial Vehicle (UAV) technology is increasingly gaining attention in localization systems due to its flexibility and mobility. However, traditional localization techniques often fail in complex environments where line-of-sight paths are obstructed. To address this challenge, this paper presents an innovative UAV-assisted high-precision multi-target localization system. The system utilizes UAVs equipped with Reconfigurable Intelligent Surfaces to create a reflective signal path, allowing a receiver sensor to capture these signals, creating favorable conditions for multi-target localization. Exploiting the sparsity of signals, we introduce a direct positioning algorithm that leverages Atomic Norm Minimization (ANM) to estimate the target’s location. To address the high complexity of traditional ANM methods, we propose a novel Coyote-ANM-based direct localization (CADL) approach. This method combines the coyote optimization algorithm with the alternating direction method of multipliers to achieve high-accuracy positioning with reduced computational complexity. Simulation results across various signal-to-noise ratio scenarios demonstrate that the proposed algorithm significantly improves localization accuracy, achieving lower root mean square error values and faster execution times compared to traditional methods.

Environmental Condition Aware Super-Resolution Acceleration Framework in Server-Client Hierarchies

In the current landscape, high-resolution (HR) videos have gained immense popularity, promising an elevated viewing experience. Recent research has demonstrated that the video super-resolution (SR) algorithm, empowered by deep neural networks (DNNs), can substantially enhance the quality of HR videos by processing low-resolution (LR) frames. However, the existing DNN models demand significant computational resources, posing challenges for the deployment of SR algorithms on client devices. While numerous accelerators have proposed solutions, their primary focus remains on client-side optimization. In contrast, our research recognizes that the HR video is originally stored in the cloud server and presents an untapped opportunity for achieving both high accuracy and performance improvements. Building on this insight, this article introduces an end-to-end video CODEC-assisted super-resolution (E 2 SR+) algorithm, which tightly integrates the cloud server with the client device to deliver a seamless and real-time video viewing experience. We propose the motion vector search algorithm executed in a cloud server, which can search the motion vectors and residuals for a part of the HR video frames and then pack them as add-ons. We also design an auto-encoder algorithm to down-sample the residuals to save the bitstream cost while guaranteeing the quality of the residuals. Lastly, we propose a reconstruction algorithm performed in the client to quickly reconstruct the corresponding HR frames using the add-ons to skip part of the DNN computations. To implement the E 2 SR+ algorithm, we design corresponding E 2 SR+ architecture in the client, which achieves significant speedup with minimal hardware overhead. Given that the environmental condition varies in the server–client hierarchies, we believe that simply applying E 2 SR+ to all frames is irrational. Accordingly, we offer an environmental condition–aware system to chase the best performance while adapting to the diverse environment. In the system, we design a linear programming (LP) model to simulate the environment and allocate frames to three existing mechanisms. Our experimental results demonstrate that the E 2 SR+ algorithm enhances the peak signal-to-noise ratio by 1.2, 2.5, and 2.3 compared with the state-of-the-art (SOTA) methods EDVR, BasicVSR, and BasicVSR++, respectively. In terms of performance, the E 2 SR+ architecture offers significant improvements over existing SOTA methods. For instance, while BasicVSR++ requires 98 ms on an NVIDIA V100 graphics processing unit (GPU) to generate a 1,280 × 720 HR frame, the E 2 SR+ architecture reduces the execution time to just 39 ms, highlighting the efficiency and effectiveness of our proposed method. Overall, the E 2 SR+ architecture respectively achieves 1.4×, 2.2×, 4.6×, and 442.0× performance improvement compared with ADAS, ISRAcc, the NVIDIA V100 GPU, and a central processing unit. Lastly, the proposed system showcases its superiority and surpasses all the existing mechanisms in terms of execution time when varying environmental conditions.

Spinal navigation with AI-driven 3D-reconstruction of fluoroscopy images: an ex-vivo feasibility study

BackgroundWith the increasing number of surgeries utilizing spinal instrumentation, three-dimensional surgical navigation aims to improve the accuracy of implant placement. However, its widespread clinical adaption has been hindered by factors such as high radiation exposure and interference with standard surgical workflows.MethodsX23D is a novel AI-based fluoroscopy reconstruction technique that generates a 3D anatomical model of the spine from only four fluoroscopy images. Based on this technology, we developed a prototype for the surgical navigation of pedicle screws placement of the lumbar spine, visualizing the 3D-reconstructed spine anatomy and the surgical drill position in real-time. An ex-vivo study was conducted to compare the accuracy of the X23D-based navigation approach with fluoroscopy-aided freehand instrumentation. Five board-certified surgeons placed pedicle screws on six human torsi within a realistic surgical environment. Breach rate, site and extent (Gertzbein-Robbins) were evaluated in postoperative CT scans, as well as execution time, radiation dose, and user experience. Specimens, operating side, and surgeon were randomised.ResultsForty-nine pedicle screws (n = 24 × 23D, n = 25 2D-fluoroscopy) were evaluated, with six breaches occurring in the control group, one of which was considered clinically significant (medial breach grade C). Five breaches with one clinically significant breach were observed in the X23D group. Breach rate, execution time for each lumbar level (X23D 167 s vs. control 156 s), radiation dose (X23D 33.26 mGy vs. control 49.5 mGy), and user experience did not reveal significant differences (p > 0.05) between the groups.ConclusionsSpinal navigation using the X23D-based approach shows promise and performs well in a realistic surgical ex-vivo setting. With further refinements, its accuracy is expected to match clinical-grade navigation systems while reducing radiation dose.

Mathematical model based on nonlinear differential equations and its control algorithm

Abstract A practical examination of the traditional robotic arm (RA) in operation revealed a significant limitation in its ability to control the position of motion. This underscores the urgent need to enhance the current RA’s position control capabilities. Therefore, this study proposes the use of nonlinear differential equations (NDEs) to establish a mathematical model and the design of NDE-based RA motion control algorithms in conjunction with a central pattern generator neural network. A comparison of the control effects showed that the proposed method was highly fitted to the target trajectory. The joint node (JN) motion tracking trajectories of the three RAs were similar, up to 90–85% to the target trajectories of the JNs. In addition, the control of the motion position was similar up to 95–98% to the target motion position trajectories. The motion control algorithm based on NDEs was effective in improving the average execution time of the Pareto optimal frontier of the RA by 58.29%. The joint velocity and angle changes of the three types of RAs under the NDE control algorithm exhibited a high degree of similarity to the fluctuations observed in the expected and predicted curves. These observations contribute to an understanding of the effectiveness of the system observer in observing the joint angle changes. This indicates that the motion control based on NDEs can effectively enhance the tracking effectiveness of the JN positions of the RA, improve the control ability of the RA motion, and increase the joint stability of the RA.

An efficient single-shot global localization method for indoor mobile robots usingphase correlation

Purpose The purposes of this paper are to solve the low-efficiency problem caused by large search space in global localization and develop an efficient global localization method requiring only one 2D-LiDAR scan to match against the prior map for indoor mobile robots. Design/methodology/approach This paper solves the global localization problem using phase correlation as the underlying registration method. To obtain accurate rotation parameter, this paper exhaustively pre-rotates the prior map by a certain angle stride in advance. Then the input scan is matched against the pre-rotated maps one by one using phase correlation to determine translation parameters, and this paper constructs an orientation histogram by the correlation coefficients. The map rotation angle and corresponding translation parameters of the maximal peak value in the orientation histogram constitute the global pose. This paper applies a divide-and-conquer method to reduce the time consumption of single phase correlation and determines promising angle ranges where the maximal peak value may appear based on the periodicity of 90º in the orientation histogram with the signal-to-noise ratio (SNR) to reduce execution times of phase correlation. Findings Both simulated and real experimental results reveal that the proposed method achieves a high enough success rate and efficient (processing time in a second) global localization. Originality/value The proposed method constructs an orientation histogram to improve the global localization success rate and applies a divide-and-conquer method with SNR to improve efficiency, which will benefit the indoor mobile robots equipped with 2D-LiDAR.

GPU Acceleration of the Boys Function Evaluation in Computational Quantum Chemistry

ABSTRACTThe Boys function, a mathematical integral function, plays a pivotal role and is frequently evaluated in ab initio molecular orbital computations. The main contribution of this paper is to accelerate the bulk evaluation of the Boys function through the effective utilization of GPUs. The proposed GPU implementation addresses GPU‐specific programming issues such as warp divergence and coalesced/stride access to global memory, and we employ the optimal numerical evaluation method from four methods based on input values to ensure efficient computation with sufficient accuracy. Moreover, to consider actual computation of molecular integrals, we have implemented and evaluated the proposed method in two scenarios: single evaluation, which computes a single value of the Boys function for a single input, and incremental evaluation, which computes multiple values of the Boys function incrementally. The execution time of the proposed GPU implementation was evaluated for both scenarios using an NVIDIA A100 Tensor Core GPU. As a result, the GPU‐accelerated bulk evaluation has achieved a throughput of computing the values of the Boys function times per second for the single evaluation and times per second for the incremental evaluation, respectively. Our parallelized CPU and GPU implementation is available at https://github.com/sstsuji/Boys‐function‐GPU‐library.