Introduction to the Special Issue on Distributed Computing with Applications in Bioengineering

Real-time monitoring microblog data can find sensitive information in time and provide help for public sentiment management and control. However, it needs processing large-scale data stream. MapReduce is a framework of processing large-scale data in batch mode, its purpose is to increase throughput, but its real-time performance is limited. Aiming at the real-time performance limitation of MapReduce, RT-SSP Real-Time Staged Stream Processing, a hybrid staged real-time stream processing scheme both for batch and real-time processing was proposed. By this method large-scale high-speed data stream is locally processed in stages, the communication cost is reduced by storing intermediate results to local node, and key technologies such as cache optimization are used to realize high concurrent read and write. Experiments show that RT-SSP scheme can improve the real-time performance of processing large-scale microblog data stream and achieve speed-up ratio of about 2.3.

Matrix factorization (MF), as one of the most accurate and scalable approaches in dimension reduction techniques, has become popular in the collaborative filtering (CF) recommender systems, social network and graph communities. Currently, Kullback-Lerbler Non-negative Matrix Factorization (KL-NMF) is one of the most famous approaches for MF, due to its representative non-negativity feature of the CF model. However, it is non-trivial to obtain high performance KL-NMF on Graphic Processing Units (GPU) for large-scale problems, due to the redundant large-scale intermediate data, frequent matrices manipulation, and access of sparse and irregular entries characteristic of KL-NMF. In this work, we propose single-thread-based KL-NMF, which depends on the involved feature tuples multiplication and summation, and then, we present L <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> norm regularized single-thread-based KL-NMF. On that basis, a novel CUDA parallelization KL-NMF approach (CuKL-NMF) is presented for GPU computing. Furthermore, to process large-scale CF data sets and make advantages of GPU computation power, we propose multi-GPU CuKL-NMF (MCuKL-NMF). Compared with state-of-the-art parallel algorithms, e.g., CUMF, CCD++, MCuKL-NMF obtains the highest performance.

Data partitioning algorithms aiming to minimize the execution time and the energy of computations in self-adaptable data-parallel applications on modern extreme-scale multicore platforms must address two critical challenges. First, they must take into account the new complexities inherent in these platforms such as severe resource contention and non-uniform memory access. Second, they must have low practical runtime and memory costs. The sequential data partitioning algorithms addressing the first challenge have a theoretical time complexity of O( $m * m * p * p$ ) where $m$ is the number of points in the discrete speed/energy function and $p$ is the number of available processors. They, however, exhibit high practical runtime cost and excessive memory footprint, therefore, rendering them impracticable for employment in self-adaptable applications executing on extreme-scale multicore platforms. We present, in this paper, the parallel data partitioning algorithms that address both the challenges. They take as input the functional models of performance and energy consumption against problem size and output workload distributions, which are globally optimal solutions. They have a low time complexity of O( $m * m * p$ ) thereby providing a linear speedup of O( $p$ ) and low memory complexity of O( $n$ ) where $n$ is the workload size expressed as a multiple of granularity. They employ dynamic programming approach, which also facilitates the easier integration of performance and energy models of communications. We experimentally study the practical cost of application of our algorithms in two data-parallel applications, matrix multiplication and fast Fourier transform, on a cluster in Grid’5000 platform. We demonstrate that their practical runtime and memory costs are low making them ideal for employment in self-adaptable applications. We also show that the parallel algorithms exhibit tremendous speedups over the sequential algorithms. Finally, using theoretical analysis for a forecast exascale platform, we demonstrate that the parallel algorithms have negligible execution times compared to the matrix multiplication application executing on the platform.

In this era of large-scale data, distributed systems built on top of clusters of commodity hardware provide cheap and reliable storage and scalable processing of massive data. Here, we review recent work on developing and implementing randomized matrix algorithms in large-scale parallel and distributed environments. Randomized algorithms for matrix problems have received a great deal of attention in recent years, thus far typically either in theory or in machine learning applications or with implementations on a single machine. Our main focus is on the underlying theory and practical implementation of random projection and random sampling algorithms for very large very overdetermined (i.e., overconstrained) $\ell_1$ and $\ell_2$ regression problems. Randomization can be used in one of two related ways: either to construct sub-sampled problems that can be solved, exactly or approximately, with traditional numerical methods; or to construct preconditioned versions of the original full problem that are easier to solve with traditional iterative algorithms. Theoretical results demonstrate that in near input-sparsity time and with only a few passes through the data one can obtain very strong relative-error approximate solutions, with high probability. Empirical results highlight the importance of various trade-offs (e.g., between the time to construct an embedding and the conditioning quality of the embedding, between the relative importance of computation versus communication, etc.) and demonstrate that $\ell_1$ and $\ell_2$ regression problems can be solved to low, medium, or high precision in existing distributed systems on up to terabyte-sized data.

Aiming at the defects of large-scale large scale wireless sensor network data processing network traffic and high task latency, a data collection scheme of mobile node based on discrete elastic collision optimization algorithm and adaptive block compression sensing is proposed. Firstly, by analyzing the relationship between the network partitioning and the node deployment, an adaptive block compressed sensing data collection strategy is proposed to realize sensor node based on adaptive network block compressed sensing data collection. Designing mobile node data acquisition path planning strategy and multiple mobile nodes The collaborative computer system adopts the fitness value constraint transformation processing technology and the parallel discrete elastic collision optimization algorithm to achieve the purpose of balancing network node energy consumption and reducing data processing task delay. Finally, the simulation results show that the data collection scheme can effectively realize high-efficiency processing of large-scale sensor network data, and reduce network traffic and network task delay, and better balance network node energy consumption.

GeoBeam: A distributed computing framework for spatial data

Parallel system is mainly composed of parallel algorithms which are cost optimal. In this paper a parallel algorithm the hash partitioned apriori (HPA) is taken into consideration. HPA partitions the candidate itemsets among processors using a hash function, like the hash join in relational databases. HPA effectively utilizes the whole memory space of all the processors, hence it works well for large scale data mining in a parallel and distributed environment. The optimization technique of dynamic data allocation is discussed for the execution of this application. This technique is applied in a parallel and distributed environment. Writing parallel data mining algorithms in a distributed environment is a non-trivial task. The main purpose of the proposed method is to meet certain challenges associated with parallel and distributed data mining such as (i) minimizing I/O (ii) Increasing processing speed (iii) Communication cost.

Land-surface temperature (LST) is a very important parameter in the geosciences. Conventional LST retrieval is based on large-scale remote-sensing (RS) images where split-window algorithms are usually employed via a traditional stand-alone method. When using the environment to visualize images (ENVI) software to carry out LST retrieval of large time-series datasets of infrared RS images, the processing time taken for traditional stand-alone servers becomes untenable. To address this shortcoming, cluster-based parallel computing is an ideal solution. However, traditional parallel computing is mostly based on the Linux environment, while the LST algorithm developed within the ENVI interactive data language (IDL) can only be run in the Windows environment in our project. To address this problem, we combine the characteristics of LST algorithms with parallel computing, and propose the design and implementation of a parallel LST retrieval algorithm using the message-passing interface (MPI) parallel-programming model on a Windows-based PC cluster platform. Furthermore, we present our solutions to the problems associated with performance bottlenecks and fault tolerance during the deployment stage. Our results show that, by improving the parallel environment of the storage system and network, one can effectively solve the stability issues of the parallel environment for large-scale RS data processing.

With the onset of the big data age, data is growing exponentially, and the issue of how to optimize large-scale data processing is especially significant.Large-scale global optimization (LSGO) is a research topic with great interest in academia and industry.Spark is a popular cloud computing framework that can cluster large-scale data, and it can effectively support the functions of iterative calculation through resilient distributed datasets (RDD).In this paper, we propose a hybrid mechanism of particle swarm optimization (PSO) and differential evolution (DE) algorithms based on Spark (SparkPSODE).The SparkPSODE algorithm is a parallel algorithm, in which the RDD and island models are employed.The island model is used to divide the global population into several subpopulations, which are applied to reduce the computational time by corresponding to RDD's partitions.To preserve population diversity and avoid premature convergence, the evolutionary strategy of DE is integrated into SparkPSODE.Finally, SparkPSODE is conducted on a set of benchmark problems on LSGO and show that, in comparison with several algorithms, the proposed SparkPSODE algorithm obtains better optimization performance through experimental results.

Next generation sequencing (NGS) technologies have enabled cheap, large-scale, and high-throughput production of short DNA sequence reads and thereby have promoted the explosive growth of data volume. Unfortunately, the produced reads are short and prone to contain errors that are incurred during sequencing cycles. Both large data volume and sequencing errors have complicated the mapping of NGS reads onto the reference genome and have motivated the development of various aligners for very short reads, typically less than 100 base pairs (bps) in length. As read length continues to increase, propelled by advances in NGS technologies, these longer reads tend to have higher sequencing error rates and more true mutations (including substitutions, insertions, or deletions) to the genome. Such new characteristics make inefficient the aligners, which are optimized for very short reads and support only ungapped alignments or gapped alignments with very limited number of gaps (typically one gap), and thereby call for new aligners with fully gapped alignment supported. In this chapter, we present the CUSHAW software suite for NGS read alignment, which is open-source and consists of three individual aligners: CUSHAW, CUSHAW2, and CUSHAW3. This suite offers parallel and efficient NGS read alignments to large genomes, such as the human genome, by harnessing multi-core CPUs or compute unified device architecture (CUDA)-enabled graphics processing units (GPUs). Moreover, it has the capability to align both base-space and color-space reads and is consistently shown to be one of the best alignment tools through our performance evaluations.

In online learning environments, the need to monitor student engagement to achieve desired learning outcomes is foundational to any pedagogy for effective online teaching. With the absence of in-person supervision, keeping track of student attention in remote classes is quite arduous, if not impossible. This paper proposes an IoT-based EEG device for the real-time assessment of students' brain activity to measure and possibly improve students' attention levels in virtual classes. An ensemble machine learning (ML) study, specifically an artificial neural network (ANN) approach, is employed to investigate the relationship between student performance and EEG data. The study shows the performance of students is negatively correlated to the delta power and the theta/alpha ratio, the common EEG metrics for mental fatigue and drowsiness, respectively, for the student. An IoT-enabled EEG device provides teachers with real-time, precise, and unbiased data regarding the student's cognitive attention levels, via a reporting system. Thus, this research demonstrates that the proposed system, utilizing EEG and ANN ensemble-ML methods, can predict attention levels in real-time, enabling timely intervention for students who are disconnected. The study opens the door to the use of advanced BCI systems in teaching to maximize student attention.

Classification based on EEG data in an RSVP experiment is considered. Although the latency in neural response relative to the stimulus onset time may be more realistically considered to vary across trials due to factors such as subject fatigue and environmental distractions, it is nevertheless assumed to be time-locked to the stimulus in most of the existing work as a means to alleviate the computational complexity. We consider here a more practical scenario that allows variation in response latency and develop a rigorous statistical formulation for modeling the uncertainty within the varying latency coupled with a likelihood ratio test (LRT) for classification. The new model not only improves the EEG classification performance, but also may predict the true stimulus onset time when this information is not precisely available. We test the proposed LRT algorithm on an EEG data set from an image RSVP experiment and show that, by admitting the latency variation, the proposed approach consistently outperforms a method that relies on perfect time-locking (AUC: 0.88 vs 0.86), especially when the stimulus onset time is not precisely available (AUC: 0.84 vs 0.71). Furthermore, the predicted stimulus onset times are highly enriched around the true onset time with p-value = 5.2 x 10(44).

Emotions directly affect essential functions in human cognition, decision-making, and interactions. Any change in a person’s emotions can lead to different patterns in brain wave signals. Electroencephalography (EEG) is one of the key methods used to measure the brain's electrical activity. Emotion identification through EEG data has become a crucial element of human-computer interaction. Recently, the increasing need for monitoring brain activity has been measured by EEG devices, and the high price of these devices, which are not accessible outside of the laboratory for personal use, has prompted the development of low-cost wearable EEG devices. This paper introduces the design and implementation of an EEG system using available and low-cost components. This EEG system applies to emotion identification using the STM32F103C8T6 microcontroller, two-stage amplifiers, and filtering by bandwidth of (0.5-48) Hz. These data are logged and processed using MATLAB 2024, where Power Spectral Density (PSD) is applied for feature extraction and then the Support Vector Machine (SVM) for emotion detection, such as boredom, calm, fear, and happiness. Results show that this system can achieve an accuracy of a range between (79-83) % for the four emotion classifications. This study concludes by summarizing the practical importance of Electroencephalography (EEG) signals in emotion recognition, focusing on its potential for future applications.

Brain signals extracted through brain-computer interface systems (BCI2000- http://www.bci2000.org) allow researchers and computer scientists to cooperate with techniques, mathematical models and statistical inferences that allow the interpretation of a variety of signals provided by people with conditions that significantly affect the ability to move or perform motor activities due to limitations in muscles, bones or nervous system. For this study, we propose a preliminary test with the root mean square (rms) fluctuation function, with EEG data, whose task was the response given to real/imaginary motor stimulus. To validate the model and all the steps up to the configuration of the rms function, we chose the information contained in the EEG of subject S003, available in the public database https://physionet.org/content/eegmmidb/1.0.0/. Considering the distribution of electrodes in the brain (lobes: frontal, parietal, temporal and occipital) and given the data availability conditions (10 - 10 system, EDF format and 160 samples per second), we analyzed 12 of the 64 channels and four stimuli, namely: opening and closing the left or right fist, imagining opening and closing the left or right fist, opening and closing both fists or both feet and imagining opening and closing both fists or both feet. We evaluated their fluctuations individually and the amplitudes of channels 32 and 37 in relation to the others (11, 22, 24, 43, 44, 49, 54, 61, 63 and 64). We observed quantitative similarities when the brain performs the same real/imaginary motor task and that the time of the amplitude changes with the increase of the scale n (time scales). In all experiments (S003_R3, S003_R4, S003_R5, S003_R6), channel 32 x 24, for n &amp;gt; 20 (15 seconds) was smaller than the others, showing that channel 32 (left hemisphere) has the largest fluctuation. From data processing (.EDF) to visualization of F&amp;lt;SUB&amp;gt;DFA&amp;lt;/SUB&amp;gt;/∆log curves, we conclude that it is possible to replicate the study for more channels, as well as to investigate other types of activities in the human brain adapted to potential variations (DDP) generated by neurons via signals extracted from the EEG device.

The sharp increase of the amount of Internet Chinese text data has significantly prolonged the processing time of classification on these data. In order to solve this problem, this paper proposes and implements a parallel naive Bayes algorithm (PNBA) for Chinese text classification based on Spark, a parallel memory computing platform for big data. This algorithm has implemented parallel operation throughout the entire training and prediction process of naive Bayes classifier mainly by adopting the programming model of resilient distributed datasets (RDD). For comparison, a PNBA based on Hadoop is also implemented. The test results show that in the same computing environment and for the same text sets, the Spark PNBA is obviously superior to the Hadoop PNBA in terms of key indicators such as speedup ratio and scalability. Therefore, Spark-based parallel algorithms can better meet the requirement of large-scale Chinese text data mining.