Parallel Data Research Articles

STRsensor: a computationally efficient method for STR allele-typing from massively parallel sequencing data.

Short tandem repeats (STRs) represent one of the most polymorphic variations in the human genome, finding extensive applications in forensics, population genetics and medical genetics. In contrast to the traditional capillary electrophoresis (CE) method, genotyping STRs using massive parallel sequencing technology offers enhanced sensitivity and accuracy. However, current methods are mainly designed for target sequencing with higher coverage for a specific STR locus, thereby constraining the utility of STRs in low- and medium-coverage whole genome sequencing (WGS) data. Here, we introduce STRsensor, a method designed to type STR alleles in low-coverage WGS data and target sequencing data, achieving a significant high detection ratio and accuracy. STRsensor employs two methods for STR allele-typing: the Kmers-based method and the CIGAR-based method. Furthermore, by incorporating a model for PCR stutters, STRsensor greatly enhances the accuracy of STR allele typing. With simulation data, we demonstrate that STRsensor achieves a detection ratio of 100$\%$ and an accuracy of 99.37$\%$ for a 30$\times $ WGS data, outperforming the existing methods, such as STRait Razor, STRinNGS, and HipSTR. When applied to real target sequencing data from 687 individuals, STRsensor achieves a detection ratio of 99.64$\%$ and an accuracy of 99.99$\%$. Moreover, STRsensor is a computationally efficient method that runs 79 times faster than HipSTR and 10 000 times faster than STRinNGS. STRsensor is freely available on GitHub: https://github.com/ChenHuaLab/STRsensor.

Performance prediction of sintered NdFeB magnet using multi-head attention regression models

The preparation of sintered NdFeB magnets is complex, time-consuming, and costly. Data-driven machine learning methods can enhance the efficiency of material synthesis and performance optimization. Traditional machine learning models based on mathematical and statistical principles are effective for structured data and offer high interpretability. However, as the scale and dimensionality of the data increase, the computational complexity of models rises dramatically, making hyperparameter tuning more challenging. By contrast, neural network models possess strong nonlinear modeling capabilities for handling large-scale data, but their decision-making and inferential processes remain opaque. To enhance interpretability of neural network, we collected 1,200 high-quality experimental data points and developed a multi-head attention regression model by integrating an attention mechanism into the neural network. The model enables parallel data processing, accelerates both training and inference speed, and reduces reliance on feature engineering and hyperparameter tuning. The coefficients of determination for remanence and coercivity are 0.97 and 0.84, respectively. This study offers new insights into machine learning-based modeling of structure-property relationships in materials and has potential to advance the research of multimodal NdFeB magnet models.

PIMSAB: A Processing-In-Memory System with Spatially-Aware Communication and Bit-Serial-Aware Computation

Bit-serial Processing-In-Memory (PIM) is an attractive paradigm for accelerator architectures, for parallel workloads such as Deep Learning (DL), because of its capability to achieve massive data parallelism at a low area overhead and provide orders-of-magnitude data movement savings by moving computational resources closer to the data. While many PIM architectures have been proposed, improvements are needed in communicating intermediate results to consumer kernels, for communication between tiles at scale, for reduction operations, and for efficiently performing bit-serial operations with constants. We present PIMSAB, a scalable architecture that provides a spatially aware communication network for efficient intra-tile and inter-tile data movement and provides efficient computation support for generally inefficient bit-serial compute patterns. Our architecture consists of a massive hierarchical array of compute-enabled SRAMs (CRAMs), which is codesigned with a compiler to achieve high utilization. The key novelties of our architecture are (1) in providing efficient support for spatially aware communication by providing local H-tree network for reductions, by adding explicit hardware for shuffling operands, and by deploying systolic broadcasting, as well as (2) by taking advantage of the divisible nature of bit-serial computations through adaptive precision and efficient handling of constant operations. These innovations are integrated into a tensor expressions-based programming framework (including a compiler for easy programmability) that enables simple programmer control of optimizations for mapping programs into massively parallel binaries for millions of PIM processing elements. When compared against a similarly provisioned modern Tensor Core GPU (NVIDIA A100), across common DL kernels and end-to-end DL networks (Resnet18 and BERT), PIMSAB outperforms the GPU by 4.80×, and reduces energy by 3.76×. We compare PIMSAB with similarly provisioned state-of-the-art SRAM PIM (Duality Cache) and DRAM PIM (SIMDRAM), and observe a speedup of 3.7× and 3.88×, respectively.

Improve Transmission Fault Tolerance and Speed for Distributed Machine Learning

Abstract. As a matter of fact, with the advent of the big data era, traditional stand-alone machine learning methods are facing computational as well as storage bottlenecks. On this basis, distributed machine learning has become a popular research direction. Withs this in mind, this research explores the basic principles of distributed machine learning and its application on large-scale datasets. According to the analysis, it is shown that the use of distributed architecture can effectively shorten the model training time and improve the scalability of the system. By comparing different distributed algorithms, it is found that the architecture based on parameter servers has obvious advantages in dealing with heterogeneous data. In addition, the experimental results show that a reasonable data partitioning and parallel computing strategy can significantly improve the training efficiency and the final performance of the model at the same time. The significance of this study is that it provides a theoretical foundation and practical guidance for distributed machine learning, which helps to promote its wide application in industry.

A Comparative Evaluation of the Genetic Variant Spectrum in the USH2A Gene in Russian Patients with Isolated and Syndromic Forms of Retinitis Pigmentosa.

Pathogenic variants in the USH2A gene are the primary cause of both non-syndromic autosomal recessive inherited retinitis pigmentosa (RP) and the syndromic form, characterized by retinal degeneration and sensorineural hearing loss. This study presents a comparative assessment of the genetic variant spectrum in the USH2A gene among Russian patients in two clinical groups. A retrospective analysis was conducted on massive parallel panel sequencing data from 2415 blood samples of unrelated patients suspected of having hereditary retinal diseases. The copy number of USH2A exons was determined using the quantitative MLPA method with the MRC-Holland SALSA MLPA kit. Biallelic pathogenic and likely pathogenic variants in the USH2A gene were identified in 69 patients (8.7%). In the group of patients with isolated hereditary RP (55 patients), the most frequent pathogenic variants were p.(Glu4445_Ser4449delinsAspLeu) (20.9%), p.(Trp3955*) (15.5%), and p.(Cys934Trp) (5.5%). In patients with the syndromic form (14 patients), the most frequent variants were p.(Trp3955*) (35.7%) and c.8682-9A>G (17.9%). It was found that patients with isolated vision impairment rarely had two "null" variants (17.8%), whereas this was common among patients with both hearing and vision impairment (71.4%) (p ≤ 0.05), explaining the severity of the disease and the earlier onset of clinical symptoms in the syndromic form of RP. Ten previously undescribed loss-of-function variants were identified. The estimated prevalence of USH2A-associated retinal dystrophy in Russia was 1.9 per 100,000 individuals. The obtained data on the differences in the spectra of genetic variants in the USH2A gene in the two studied groups highlight the importance of establishing genotype-phenotype correlations and predicting disease severity, aiming at potential early cochlear implantation and selection of target therapy.

Efficient Cross-Domain Fault Diagnosis via Distributed Multi-Source Domain Deep Transfer Learning

Abstract In the actual industrial production, the working conditions of rotating machinery are complex and changeable, and the health-state monitoring data are increasingly large and difficult to be labeled, which will seriously restrict the accuracy and efficiency of cross-domain fault diagnosis (CDFD) of rotating machinery. Therefore, an efficient multi-source domain deep transfer learning (MDDTL) method for CDFD of rotating machinery is proposed. Firstly, a MDDTL model is constructed to improve the accuracy of CDFD. In the model, a dual-phase domain alignment strategy is designed, which considers the alignment of feature distributions between each source and target domain pair in the feature space and that of prediction probabilities between domain-specific fault classifiers in the output space. The fault prediction results from multiple different fault classifiers are merged dynamically by the proposed imbalanced adaptive prediction strategy. Secondly, a data-parallel distributed training scheme for MDDTL model is proposed. Based on the idea of data parallelism, the distributed parallel training of MDDTL model is performed with Horovod-GPU platform, and the parameters are synchronously updated with the bandwidth-optimal Ring-AllReduce architecture. Under the premise of ensuring the accuracy of fault diagnosis (FD), the training time of MDDTL model is significantly reduced. Finally, extensive experiments are conducted to verify the effectiveness of the proposed MDDTL method. The results demonstrate that the proposed method not only effectively improves the accuracy of CDFD of rotating machinery, but also significantly improves the training efficiency of MDDTL model. After adopting the proposed method, the diagnosis accuracies achieved under two different cross-working condition scenarios reach 97.09% and 97.87% respectively, and the model training time is reduced by 73.62% when facing a large-scale rotating machinery training set.

Enhancing computational efficiency in topology-optimized mode converters via dynamic update rate strategies

In the big data era, mode division multiplexing, as a technology for extended channel capacity, demonstrates potential in enhancing parallel data processing capability. Consequently, developing a compact, high-performance mode converter through efficient design methods is an urgent requirement. However, traditional design methodologies for these converters face significant computational complexities and inefficiencies. Addressing this challenge, this paper introduces a novel topology optimization design method for mode converters employing a Dynamic Adjustment of Update Rate (DAUR). This approach markedly reduces computational overhead, accelerating the design process while ensuring high performance and compactness. As a proof-of-concept, an ultra-compact dual-mode converter was designed. The DAUR method demonstrated an 80% reduction in computational time compared to traditional methods, while maintaining a compact design (only 1.4 μm × 1.4 μm) and an insertion loss under 0.68 dB across a wavelength range of 1525 nm to 1575 nm. Meanwhile, simulated inter-mode crosstalk remained below − 24 dB across a 40 nm bandwidth. A comprehensive comparison with traditional inverse design algorithms is presented, demonstrating our method’s superior efficiency and effectiveness. Our findings suggest that DAUR not only streamlines the design process but also facilitates exploration into more complex micro-nano photonic structures with reduced resource investment.

Physics-informed spatio-temporal hybrid neural networks for predicting remaining useful life in aircraft engine

Aircraft engines have complex structures and various operating conditions, which leads to the health monitoring data of aircraft engines’ high coupling in the spatial domain and time invariance in the temporal domain. Existing methods for remaining useful life prediction of aircraft engines are deficient in handling spatial data and spatial–temporal data fusion. To address this problem, a novel method named physics-informed spatio-temporal hybrid neural networks (PI-STHNN) is proposed. In this study, a spatial domain data association graph for aircraft engines is constructed by utilizing a thermal cycle physical model as prior knowledge and combining it with the grey relational grade analysis method. Specifically, the proposed spatio-temporal hybrid neural network framework extracts the temporal features and spatial coupling features of the data in parallel. Then attention mechanism is employed to weigh and fuse all feature representations. Finally, the method is validated on the two aircraft engine datasets, the results show that PI-STHNN outperforms current state-of-the-art methods. Furthermore, our analysis sheds light on the advantages of incorporating physical information into graph neural network models and visually presents the attention weights to elucidate the contributions of time-domain and spatial-domain models within our hybrid approach.

Design and implementation of a nondestructive testing system for magnetic field imaging based on machine learning

This study presents the design of a nondestructive testing (NDT) system for detecting metal defects using a magnetic field sensor array. The system integrates a hardware-software detection framework that includes parallel computation and processing cores, along with multiple anisotropic magnetoresistance (AMR) sensors. These AMR sensors capture subtle variations in the magnetic field, which serve as the foundation for defect identification. The system enables synchronized data acquisition from multiple AMR sensors, achieving multi-axis data fusion and parallel signal processing. Furthermore, the detection system gathers defect features to train an NDT system based on machine learning (ML). Subsequently, ML-generated defect imaging features are processed through an imaging framework, producing magnetic field imaging (MFI) for precise defect localization. The study tested six different metal samples with various defect sizes and types to validate the system. The results demonstrate the system’s ability to accurately detect and identify different defects, highlighting the high precision of magnetic field detection and the overall effectiveness of the proposed system for NDT.

SPSC: Stream Processing Framework Atop Serverless Computing for Industrial Big Data.

With the advance of smart manufacturing and information technologies, the volume of data to process is increasing accordingly. Current solutions for big data processing resort to distributed stream processing systems, such as Apache Flink and Spark. However, such frameworks face challenges of resource underutilization and high latency in big data application scenarios. In this article, we propose SPSC, a serverless-based stream computing framework where events are discretized into the atomic stream and stateless Lambda functions are taken as context-irrelevant operators, achieving task parallelism and inherent data parallelism in processing. Also, we implement a prototype of the framework on Amazon Web service (AWS) using AWS Lambda, AWS simple queue service, and AWS DynamoDB. The evaluation shows that compared with Alibaba's real-time computing Flink version, SPSC outperforms by 10.12% when the overhead is close.

Multi-step reconstruction of radio-interferometric images

The large aperture arrays for the currently under construction SKA Observatory (SKAO) will allow for observations of the universe in the radio spectrum at unprecedented resolution and sensitivity. However, these telescopes will produce data on the scale of exabytes, introducing a slew of hardware and software design challenges. This paper proposes a multi-step image reconstruction framework that allows for partitioning of visibility data by baseline length. This enables more flexible data distribution and parallelization, aiding in processing radio-astronomical observations within given constraints. Additionally, as each step of the framework only relies on a subset of the total visibilities, one can perform reconstruction progressively, with the initial step performed on the SKAO Science Data Processors and the second on local clusters. The multi-step reconstruction is separated into two steps. First a low-resolution image is reconstructed with only short-baseline visibilities, and then using this image together with the long-baseline visibilities, the full-resolution image is reconstructed. The proposed method only operates in the minor cycle, and it can be easily integrated into existing imaging pipelines. We show that our proposed method allows for partitioning of visibilities by baseline without introducing significant additional drawbacks, reconstructing images of similar quality within similar numbers of major cycles compared to a single-step all-baselines approach that uses the same reconstruction method as well as compared to multi-scale CLEAN.

DeepZoning: Re-accelerate CNN Inference with Zoning Graph for Heterogeneous Edge Cluster

Parallelizing CNN inference on heterogeneous edge clusters with data parallelism has gained popularity as a way to meet real-time requirements without sacrificing model accuracy. However, existing algorithms struggle to find optimal parallel granularity for complex CNNS, the structure of which is a directed acyclic graph (DAG) rather than a chain, and the parallel dimension is inflexible. To distribute the workload of modern CNNs on heterogeneous devices is also proven as NP-hard problem. In this paper, we introduce DeepZoning , a versatile and cooperative inference framework that combines both model and data parallelism to accelerate CNN inference. DeepZoning employs two algorithms at different levels: (1) a low-level Adaptive Workload Partition algorithm that uses linear programming and takes spatial and channel dimensions into optimization during the search for feature map distribution on heterogeneous devices, and (2) a high-level Model Partition algorithm that finds the optimal model granularity and organizes complex CNNs into sequential zones to balance communication and computation during execution. Our experimental evaluations show that DeepZoning is effective, achieving up to a 3.02 × speed improvement on our experimental prototype compared to state-of-the-art algorithms.

GPU-Accelerated, Interval-Based Parameter Identification Methods Illustrated Using the Two-Compartment Problem

Interval methods are helpful in the context of scientific computing for reliable treatment of problems with bounded uncertainty. Most traditional interval algorithms, however, were designed for sequential execution while internally depending on processor-specific instructions for directed rounding. Nowadays, many-core processors and dedicated hardware for massively parallel data processing have become the de facto standard for high-performance computers. Interval libraries have yet to adapt to this heterogeneous computing paradigm. In this article, we investigate the parallelization of interval methods with an emphasis on modern graphics processors. Using a parameter identification scenario in combination with newly developed or enhanced GPU-based interval software, we evaluate different methods for reducing the size of large interval search domains. For the first time, algorithmic differentiation can be used with intervals on the GPU. Different versions of interval optimization algorithms are compared wrt. their functionality, run times, and energy consumption.

Parallel molecular data storage by printing epigenetic bits on DNA

DNA storage has shown potential to transcend current silicon-based data storage technologies in storage density, longevity and energy consumption1–3. However, writing large-scale data directly into DNA sequences by de novo synthesis remains uneconomical in time and cost4. We present an alternative, parallel strategy that enables the writing of arbitrary data on DNA using premade nucleic acids. Through self-assembly guided enzymatic methylation, epigenetic modifications, as information bits, can be introduced precisely onto universal DNA templates to enact molecular movable-type printing. By programming with a finite set of 700 DNA movable types and five templates, we achieved the synthesis-free writing of approximately 275,000 bits on an automated platform with 350 bits written per reaction. The data encoded in complex epigenetic patterns were retrieved high-throughput by nanopore sequencing, and algorithms were developed to finely resolve 240 modification patterns per sequencing reaction. With the epigenetic information bits framework, distributed and bespoke DNA storage was implemented by 60 volunteers lacking professional biolab experience. Our framework presents a new modality of DNA data storage that is parallel, programmable, stable and scalable. Such an unconventional modality opens up avenues towards practical data storage and dual-mode data functions in biomolecular systems.

Quantum classical hybrid convolutional neural networks for breast cancer diagnosis.

The World Health Organization states that early diagnosis is essential to increasing the cure rate for breast cancer, which poses a danger to women's health worldwide. However, the efficacy and cost limitations of conventional diagnostic techniques increase the possibility of misdiagnosis. In this work, we present a quantum hybrid classical convolutional neural network (QCCNN) based breast cancer diagnosis approach with the goal of utilizing quantum computing's high-dimensional data processing power and parallelism to increase diagnosis efficiency and accuracy. When working with large-scale and complicated datasets, classical convolutional neural network (CNN) and other machine learning techniques generally demand a large amount of computational resources and time. Their restricted capacity for generalization makes it challenging to maintain consistent performance across multiple data sets. To address these issues, this paper adds a quantum convolutional layer to the classical convolutional neural network to take advantage of quantum computing to improve learning efficiency and processing speed. Simulation experiments on three breast cancer datasets, GBSG, SEER and WDBC, validate the robustness and generalization of QCCNN and significantly outperform CNN and logistic regression models in classification accuracy. This study not only provides a novel method for breast cancer diagnosis but also achieves a breakthrough in breast cancer diagnosis and promotes the development of medical diagnostic technology.

The station system measures the water level and amount of rain that can save the data and online through Web Server

This research paper, the aim is to create a water level and rain gauge station system that can record data and online through Web Server. To be monitor the water level of the river and the amount of water in the rainy season easily by looking through the website. The design of the water level measurement system and the amount of rain is measured with a hydrostatic sensor, Tipping Bucket. A small measuring station, easy to install, uses solar energy and has a Data Logger to record data in parallel with an online database. In the research process, the principle of correlation is used to find the relationship between two variables to find the error of the sensor, namely the water level sensor (Hydrostatic) and the rain sensor (Tipping Bucket). Through research, it was found that the measuring station, the database system can work correctly and with high stability, the water level measurement sensor is highly accurate, the error value is between 0.01 - 0.36 m, the accuracy rate is 97.38%. 98.97% compared to water level measurement stations in nearby areas. For the rain sensor, from the actual count compared to the count with the program 3 times, it is seen that there is 1 error with a deviation value of 0.3 mm. From the recorded data of 2 months (July and August) it was found that the water level is still low to moderate, away from the dangerous water level of 6.84 meters and the amount of rain is still quite high with the amount of rain. The days account for 64.51% and 54.83% of July and August respectively, but the amount of rain will be moderate to moderate. The speed used to send data from the measuring station to the web server via 3G Internet will be between 4 – 6 seconds.

Enhancing Neural Machine Translation Quality for Kannada–Tulu Language Pairs through Transformer Architecture: A Linguistic Feature Integration

The rise of intelligent systems demands good machine translation models that are less data hungry and more efficient, especially for low- and extremely-low-resource languages with few or no data available. By integrating a linguistic feature to enhance the quality of translation, we have developed a generic Neural Machine Translation (NMT) model for Kannada–Tulu language pairs. The NMT model uses Transformer architecture and a state-of-the-art model for translating text from Kannada to Tulu and learns based on the parallel data. Kannada and Tulu are both low-resource Dravidian languages, with Tulu recognised as an extremely-low-resource language. Dravidian languages are morphologically rich and are highly agglutinative in nature and there exist only a few NMT models for Kannada–Tulu language pairs. They exhibit poor translation scores as they fail to capture the linguistic features of the language. The proposed generic approach can benefit other low-resource Indic languages that have smaller parallel corpora for NMT tasks. Evaluation metrics like Bilingual Evaluation Understudy (BLEU), character-level F-score (chrF) and Word Error Rate (WER) are considered to obtain the improved translation scores for the linguistic-feature-embedded NMT model. These results hold promise for further experimentation with other low- and extremely-low-resource language pairs.

A certain examination on heterogeneous systolic array (HSA) design for deep learning accelerations with low power computations

Acceleration techniques play a crucial role in enhancing the performance of modern high-speed computations, especially in Deep Learning (DL) applications where the speed is of utmost importance. One essential component in this context is the Systolic Array (SA), which effectively handles computational tasks and data processing in a rhythmic manner. Google's Tensor Processing Unit (TPU) leverages the power of SA for neural networks. The core SA's functionality and performance lies in the Computation Element (CE), which facilitates parallel data flow. In our article, we introduce a novel approach called Proposed Systolic Array (PSA), which is implemented on the CE and further enhanced with a modified Hybrid Kogge Stone adder (MHA). This design incorporates principles to expedite computations by rounding and extracting data model in SA as PSA-MHA. The PSA, utilizing a data flow model with MHA, significantly accelerates data shifts and control passes in execution cycles. We validated our approach through simulations on the Cadence Virtuoso platform using 65 nm process technology, comparing it to the General Matrix Multiplication (GMMN) benchmark. The results showed remarkable improvements in the CE, with a 30.29 % reduction in delay, a 23.07 % reduction in area, and an 11.87 % reduction in power consumption. The PSA outperformed these improvements, achieving a 46.38 % reduction in delay, a 7.58 % reduction in area, and an impressive 48.23 % decrease in Area Delay Product (ADP). To further substantiate our findings, we applied the PSA-based approach to pre-trained hybrid Convolutional and Recurrent (CNN-RNN) neural models. The PSA-based hybrid model incorporates 189 million Multiply-Accumulate (MAC) units, resulting in a weighted mean architecture value of 784.80 for the RNN component. We also explored variations in bit width, which led to delay reductions ranging from 20.17 % to 30.29 %, area variations between 13.08 % and 32.16 %, and power consumption changes spanning from 11.88 % to 20.42 %.

Design of an FPGA-based short-circuit fault diagnosis algorithm for DC current-limiting fuses

Abstract In recent years, DC current-limiting fuses have played a crucial role in the comprehensive power systems of ships, rail transport, and new energy sectors, ensuring the safe and reliable operation of DC power systems. However, with the increase in the capacity of DC power systems, short-circuit faults can lead to a rise in short-circuit current at rates of up to 40 A/μs, with peak short-circuit currents reaching up to 105KA. This poses higher demands on the accuracy and speed of fuse diagnostics to interrupt short-circuit currents. To address this issue, this paper introduces a fault diagnostic algorithm based on FPGA that can calculate the rise rate of short-circuit currents in real-time: employing FPGA with its high integration, abundant logic resources, and parallel data processing capabilities as the core of the measurement and control system to enhance the speed of short-circuit fault diagnosis from a hardware perspective; replacing the common large current trip protection and DDL protection algorithms with the Newton interpolation algorithm. This improvement calculates the rise rate of each current data point instead of the average rise rate over a period, thus enhancing the speed and accuracy of short-circuit fault diagnosis from a software perspective. The paper also discusses the design of the system’s hardware and software and the construction of the experimental platform. Simulation and experimental results show that the fault diagnosis time of the algorithm has been reduced from 98μs to 36μs, with the numerical error in the short-circuit current rise rate within 2%, and a significant reduction in the peak value of short-circuit current, meeting performance requirements and effectively ensuring the safe and reliable operation of DC power systems.

Munkaerő-piaci mutatók elemzése az Észak-Alföld régióban 2018 és 2022-es években

In my study, I analyze the labor market indicators of the North Great Plain statistical region for the years 2018 and 2022 in order to map the regional labour market characteristics based on the indicators. One method of the analysis is the Beveridge curve. This complex analysis method graphically illustrates the evolution of the relationship between the unemployment rate and the proportion of vacant positions typical of the region under investigation. The results of the analysis can draw attention to possible problems in the labour market in the region. I examine the secondary statistical data in parallel with the available related literature.