PLANC

Sequential pattern mining is applicable in a wide range of applications since many types of data sets are in a time related format. Besides mining sequential patterns in a single dimension, mining multidimensional sequential patterns can give us more informative and useful patterns. Due to the huge increase in data volume and also quite large search space, efficient solutions for finding patterns in multidimensional sequence data are nowadays very important. For this reason, developing a parallel algorithm is necessary. In this paper, we present a multidimensional sequence model and a parallel algorithm follows the level-wise approach and all participating processors or workers generate candidate sequence and count their supports independently. Simulation experiments show good load balancing and scalable and acceptable speedup over different processors and problem sizes. dense data sets such as telecommunications, where there are many and long frequent patterns, performance of these algorithms degrades incredibly. On the one side, parallel and distributed computing is expected to relieve current mining methods from the sequential bottleneck, providing the ability to scale to massive datasets, and improving the response time. Achieving good performance on today's multiprocessor systems is a non-trivial task. The main challenges include synchronization and communication minimization, work-load balancing, find good data layout and data decomposition, and disk I/O minimization, which is especially important for data mining (4). On the other side, the basic single dimension can not satisfy the requirement of multi-attribute analysis, which is often the case in actual system practice. To address this problem, multidimensional sequence pattern mining is developed. The most time consuming operation in the discovery process of sequential patterns is the computation of the frequency of the occurrences of interesting subsequences in the sequence database. However, the number of sequential patterns grows exponentially and the task of finding all sequential patterns requires a lot of computational resources, which make it an ideal candidate for parallel processing. However, sequential pattern mining suffers from a scalability problem in both memory use and computational time when a dataset size is large. To perform mining on large datasets, we propose to parallelize this task. We demonstrate that our parallel algorithm can effectively alleviate the scalability problem. In this paper, we present a model for multidimensional sequence pattern and then propose a parallel algorithm for mining sequential patterns from multidimensional sequence data. The algorithm is based on the data parallelism strategy of pattern growth. The experimental results show that our parallel algorithms usually achieve good load balancing and scalability.

Nonconvex low-rank tensor approximation with graph and consistent regularizations for multi-view subspace learning

The all-pairs shortest path problem is a classic problem to study characteristics of the given graphs. Though many efficient all-pairs shortest path algorithms have been published, it is still a very expensive computing task, especially with large graph datasets. In this paper, we propose an efficient parallel all-pairs shortest path algorithm based on Peng et al.'s fast sequential algorithm on shared-memory parallel environments to achieve faster and more efficient calculation for large-scale real-world networks. Peng et al.'s algorithm needs to sort vertices with respect to their degrees. However, it turns out the original algorithm uses less efficient sorting method, which is a significant portion of parallel overhead. Therefore, we also propose an efficient parallel method to sort data within a fixed range, in order to minimize the parallel overhead in our parallel algorithm. The optimized efficient sorting method can be used for general sorting purposes. Our experimental analysis shows that our proposed parallel algorithm achieves very high parallel speedup, even hyper-linear speedup, with real-world test datasets on two different shared-memory multi-core systems.

Various sensor network measurement studies have reported instances of transient faults in sensor readings. In this work, we seek to answer a simple question: How often are such faults observed in real deployments? To do this, we first explore and characterize three qualitatively different classes of fault detection methods. Rule-based methods leverage domain knowledge to develop heuristic rules for detecting and identifying faults. Estimation methods predict "normal" sensor behavior by leveraging sensor correlations, flagging anomalous sensor readings as faults. Finally, learning-based methods are trained to statistically identify classes of faults. We find that these three classes of methods sit at different points on the accuracy/robustness spectrum. Rule-based methods can be highly accurate, but their accuracy depends critically on the choice of parameters. Learning methods can be cumbersome, but can accurately detect and classify faults. Estimation methods are accurate, but cannot classify faults. We apply these techniques to four real-world sensor data sets and find that the prevalence of faults as well as their type varies with data sets. All three methods are qualitatively consistent in identifying sensor faults in real world data sets, lending credence to our observations. Our work is a first-step towards automated on-line fault detection and classification.

Reservoir modeling and simulation are vital components of modern reservoir management processes. Despite the advances in computing power and the advent of smart technologies, the implementation of model‐based operational/control strategies has been limited by the inherent complexity of reservoir models. Thus, reduce order models that not only reduce the cost of the implementation but also provide geological consistent prediction are essential. This article introduces reduced‐order models based on the proper orthogonal decomposition (POD) coupled with linear interpolation for upscaling. First, using POD‐based models, low rank approximate (LRA) are obtained by projecting the high dimensional permeability dataset to a low dimensional subspace spanned by its trajectories to decorrelate the dataset. Next, the LRA is integrated into the interpolation algorithm to generate upscaled values. This technique is highly scalable since low‐rank approximations are achieved by the variation in the number of modes used for reconstruction. To test the validity and reliability of the model, we show its application to the practical dataset from SPE10 benchmark case2. From statistics of the error analysis, the classical POD algorithm seems to be more preferred for LRA; however, since non‐negativity of the permeability data set is a priority, the constrained POD (non‐negative POD) algorithm described in this article is more appropriate. Finally, we compared the POD‐based models to a traditional industry‐standard upscaling technique (e.g., arithmetic mean) to highlight our model benefits/performance. Results show that the POD‐based models, particularly the non‐negative POD model, produce considerably less error than the arithmetic mean model in the upscaling process.

MrBayes is a popular bioinformatics software that is widely used in phylogenetic analysis. The core algorithm of Mrbayes is Metropolis Coupled Markov Chain Monte Carlo (MC3). However, when dealing with large data sets, MC3 algorithm is too slow to meet researcher's requirements. Although several parallelizations have been proposed for MrBayes, such as MPI (Message Passing Interface) based MrBayes, GPU (Graphics Processing Unit) based MrBayes, there is still no efficient parallel algorithm to fully utilize computing power of modern CPU and computer architecture. This paper (a) presents a new three-level hybrid parallel algorithm, include data-level parallelism (DLP), thread-level parallelism (TLP), and process-level parallelism (PLP), which can be used on most modern multi-core computers; (b) compares the performance of different combinations of parallel strategies on real-world protein data sets. The experimental results show that, this hybrid parallel algorithm does convert more computing powers into higher speedup. Furthermore, the proposed algorithm's speedup is near the speedup on one GPU at the same data sets. This algorithm is fit for practical use in phylogenetic inferences.

PARALLEL AND LARGE SCALE MATRIX COMPUTATIONS IN CONTROL: SOME IDEAS

Semantic feature extraction based on subspace learning with temporal constraints for acoustic event recognition


 
 
 
 Apriori algorithm is a classical algorithm of association rule mining and widely used for generating frequent item sets. However, the original Apriori algorithm has some limitation such as it needs to scan the dataset many times to discover all frequent itemset and generate huge number of candidate itemset. To overcome these limitations, researchers have made a lot of improvements to the Apriori such as candidate generation, without candidate generation, transaction reduction, partitioning, and sampling. When it comes to mine massive data, these algorithms failed to prove efficiency because limitation of the processing capacity, storage capacity, and main memory constraints. Therefore, parallel and distributed algorithms are developed to perform large-scale computing in ARM on multiple processors. However, the problems with most of the parallel and distributed framework are overheads of managing distributed system, lack of high level parallel programming language, and node failures. Hadoop-MapReduce is an efficient, scalable, and simplified programming model for massive data processing and it also available on cloud environment. Cloud computing offers huge computing resources, and capacities to solve big data challenges. Recently many parallel algorithms have been proposed on Hadoop-MapReduce to enhance the performance of Apriori algorithm but there are some drawbacks: since multiple scan over the dataset is needed to generate candidate itemset, it consume more execution time. The aim of this study is to propose a parallel Transaction Reduction MapReduce Apriori algorithm (TRMR-Apriori) which is reduce unnecessary transaction values and transactions from the dataset in parallel manner to overcome above problems. The experiments show that TRMR-Apriori is able to achieve better execution time to discover frequent itemset those of previous sequential ARM algorithms such as Apriori, AprioriTid, Eclat, and FP-Growth and the previous parallel algorithms such as PApriori, MRApriori, and Modified Apriori with different condition on homogeneous computing environment using Hadoop-MapReduce platform in cloud. Overall, the TRMR-Apriori shows the strength to extract the frequent itemset from massive dataset in cloud.
 
 
 
 

To better mine the effective information contained in massive data and improve the accuracy of short-term load forecasting, this paper proposes a hybrid model based on convolutional neural network and long short-term memory network (CNN-LSTM) for short-term load forecasting (STLF), which takes massive historical load data, meteorological data, date information, and peak-valley electric price data as input by constructing continuous feature maps with time sliding window. Convolutional neural network (CNN) is used to extract features and reshape them into vectors. The feature vectors are constructed in temporal and fed into long short-term memory network (LSTM) which is used to predict STLF. To the best of our knowledge, this is the first work to predict STLF using deep learning in both spatial and temporal domains. It is shown that the forecasting accuracy can be notably improved by CNN-LSTM hybrid model method. The effectiveness of the proposed method is validated through extensive comparison studies on a real-world dataset.

Very large spatially-referenced datasets, for example, those derived from satellite-based sensors which sample across the globe or large monitoring networks of individual sensors, are becoming increasingly common and more widely available for use in environmental decision making. In large or dense sensor networks, huge quantities of data can be collected over small time periods. In many applications the generation of maps, or predictions at specific locations, from the data in (near) real-time is crucial. Geostatistical operations such as interpolation are vital in this map-generation process and in emergency situations, the resulting predictions need to be available almost instantly, so that decision makers can make informed decisions and define risk and evacuation zones. It is also helpful when analysing data in less time critical applications, for example when interacting directly with the data for exploratory analysis, that the algorithms are responsive within a reasonable time frame. Performing geostatistical analysis on such large spatial datasets can present a number of problems, particularly in the case where maximum likelihood. Although the storage requirements only scale linearly with the number of observations in the dataset, the computational complexity in terms of memory and speed, scale quadratically and cubically respectively. Most modern commodity hardware has at least two processor cores if not more. Other mechanisms for allowing parallel computation such as Grid based systems are also becoming increasingly commonly available. However, currently there seems to be little interest in exploiting this extra processing power within the context of geostatistics. In this paper we review the existing parallel approaches for geostatistics. By recognising that different natural parallelisms exist and can be exploited depending on whether the dataset is sparsely or densely sampled with respect to the range of variation, we introduce two contrasting novel implementations of parallel algorithms based on approximating the data likelihood extending the methods of Vecchia (1988) and Tresp (2000). Using parallel maximum likelihood variogram estimation and parallel prediction algorithms we show that computational time can be significantly reduced. We demonstrate this with both sparsely sampled data and densely sampled data on a variety of architectures ranging from the common dual core processor, found in many modern desktop computers, to large multi-node super computers. To highlight the strengths and weaknesses of the different methods we employ synthetic data sets and go on to show how the methods allow maximum likelihood based inference on the exhaustive Walker Lake data set.KeywordsMessage Passing InterfacePrediction LocationApproximate Maximum LikelihoodWalker LakeMessage Passing Interface ImplementationThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Background: Sickle cell disease (SCD) is a chronic, multifaceted blood disorder that affects the oxygen-carrying protein hemoglobin (Hb). Along with chronic hemolysis and vaso-occlusion that often accompanies SCD, low Hb levels may contribute to downstream end-organ damage (EOD). Emerging evidence suggests an association between chronic hemolytic anemia in SCD and both short-term and long-term adverse outcomes, including EOD. Therefore, treatments that directly increase Hb may improve patients’ acute symptoms while reducing the risk for long-term disease-related effects known to contribute to morbidity and mortality. Aims: This study aims to ascertain the association between Hb level variation and EOD and clinical conditions typically associated with SCD (stroke, pulmonary hypertension, chronic kidney disease [CKD], end-stage renal disease [ESRD], leg ulcers, and pneumonia) in a real-world data set. Methods: The Clinical Practice Research Datalink (CPRD) and the Hospital Episode Statistics (HES) databases in England were used to retrospectively identify linked Hb data among patients with SCD (April 1, 2007, to March 31, 2019). Recurring results in any 90-day period were averaged. Logistic regression adjusted with demographic and clinical variables was used to determine odds ratios (ORs) of developing each EOD outcome or clinical complication. ORs, P values, 95% confidence intervals (CIs), and areas under the receiver operating curve (AUC-ROCs) and corresponding 95% CIs were reported. Results: Of 11,936 patients identified, 5379 (45%) had Hb results recorded in the CPRD. The logistic model for risk assessment of all outcomes studied indicated that an increase in Hb among patients with SCD was associated with a statistically significant (P<0.001) decrease in adjusted ORs per 1 g/dL increase in Hb (Table). Except for ESRD, leg ulcers, and pneumonia, the adjusted models showed statistically significant increases in AUC-ROC compared with the unadjusted models, signifying that the adjusted models were more robust and further verified a statistical OR after adjustment. Increased Hb was associated with decreased risk of respiratory-related events. Pulmonary hypertension had the strongest effect with the most robust model (ORadjusted of 0.66). For pneumonia, ORadjusted was 0.77. For renal insufficiency, ORs for developing CKD or ESRD were significantly lower than 1 (ORadjusted=0.73 and ORadjusted=0.75, respectively). After adjusting for clinical and demographic factors, the model predicted a lower risk for stroke (ORadjusted=0.89). The estimated AUC-ROCs suggest that the adjusted model is more robust in stroke, hypertension, and CKD. These findings did not change when testing for other covariates such as deep vein thrombosis. Summary-Conclusion: Among patients with SCD, an increase in Hb of 1 g/dL was associated with a statistically significant reduction in risk for 6 common EOD outcomes and clinical conditions. The results were more pronounced after adjusting for demographic and clinical covariates (based on clinical opinion and literature review) and further assessed with a bivariate analysis. The results were obtained from a representative real-world evidence dataset analyzed over a 12-year period, which was sufficient to observe EOD events, and are generalizable to the UK and similar populations. Our findings support the use of therapeutics that increase Hb in patients with SCD to protect against deleterious organ damage.

In the social computing environment, the complete information about an individual is usually distributed in heterogeneous social networks, which are presented as linked data. Synthetically recognizing and integrating these distributed and heterogeneous data for efficiently information searching is an important but challenging work. In this paper, a dynamic weight (DW)-based similarity calculation is proposed to recognize and integrate similar individuals from distributed data environments. First, each link of an individual is weighted by applying DW. Then, a semantic similarity metric is proposed to combine the DW into similarity calculation. Then, a searching system framework for a similarity-based individual is designed and tested in real-world data sets. Finally, massive experiments are conducted both in benchmark and real-world social community data sets. The results show that our approach can produce a good result in similar individual searching in social networks. In addition, it performs significantly better than the existing state-of-the-art approaches in similar individual searching.

Operator Splitting and Domain Decomposition for Multiclusters

Group trajectory pattern mining for large amounts of mobile customers is a practical task in broad applications. Usually the pattern mining result is a set of mined patterns with their support. However, most of them are not valuable for users and it is difficult for users to go through all mined patterns to find valuable ones. In this paper, instead of just mining group trajectory patterns, we investigate how to mine the top valuable patterns for users, which has not been well solved yet given the following two challenges. The first is how to estimate the value of trajectory patterns according to users’ requirements. Second, there are redundant information in the mined results because many mined patterns share common sub-patterns. To address these challenges, we define an adaptable value measuring model by leveraging multi-factors correlation in users’ requirements, which is used to estimate the value of trajectory patterns. In order to reduce the redundant sub-patterns, we propose a new group trajectory pattern mining approach directed by the adaptable value measuring model. In addition, we extend and implement the algorithm as a parallel algorithm in cloud computing platform to deal with massive data. Experiments on real massive mobile data show the effectiveness and efficiency of the proposed approach.