INTRODUCTION: The index is one of the most effective ways to improve the database query performan... more INTRODUCTION: The index is one of the most effective ways to improve the database query performance. The expert-based index recommendation approach cannot adjust the index configuration in real time. At the same time, reinforcement learning can automatically update the index and improve the recommended configuration by leveraging expert experience.OBJECTIVES: This paper proposes the RBOIRA, which combines rules and reinforcement learning to recommend the optimal index configuration for a set of workloads in a dynamic database.METHODS: Firstly, RBOIRA designed three heuristic rules for pruning index candidates. Secondly, it uses reinforcement learning to recommend the optimal index configuration for a set of workloads in the database. Finally, we conducted extensive experiments to evaluate RBOIRA using the TPC-H database benchmark.RESULTS: RBOIRA recommends index configurations with superior performance compared to the baselines we define and other reinforcement learning methods used...
Recent years have seen a surge in approaches that combine deep learning and recommendation system... more Recent years have seen a surge in approaches that combine deep learning and recommendation systems to capture user preference or item interaction evolution over time. However, the most related work only consider the sequential similarity between the items and neglects the item content feature information and the impact difference of interacted items on the next items. This paper introduces the deep bidirectional long short-term memory (LSTM) and self-attention mechanism into the sequential recommender while fusing the information of item sequences and contents. Specifically, we deal with the issues in a three-pronged attack: the improved item embedding, weight update, and the deep bidirectional LSTM preference learning. First, the user-item sequences are embedded into a low-dimensional item vector space representation via Item2vec, and the class label vectors are concatenated for each embedded item vector. Second, the embedded item vectors learn different impact weights of each item...
Airfare price prediction is one of the core facilities of the decision support system in civil av... more Airfare price prediction is one of the core facilities of the decision support system in civil aviation, which includes departure time, days of purchase in advance and flight airline. The traditional airfare price prediction system is limited by the nonlinear interrelationship of multiple factors and fails to deal with the impact of different time steps, resulting in low prediction accuracy. To address these challenges, this paper proposes a novel civil airline fare prediction system with a Multi-Attribute Dual-stage Attention (MADA) mechanism integrating different types of data extracted from the same dimension. In this method, the Seq2Seq model is used to add attention mechanisms to both the encoder and the decoder. The encoder attention mechanism extracts multi-attribute data from time series, which are optimized and filtered by the temporal attention mechanism in the decoder to capture the complex time dependence of the ticket price sequence. Extensive experiments with actual ci...
At present, some existing tag anti-collision algorithms are used mostly in scenarios of fixed num... more At present, some existing tag anti-collision algorithms are used mostly in scenarios of fixed number of tags and they perform poorly in tag moving scenarios. Therefore, this paper proposes a double reader tag movement RFID anti-collision method based on suffix, and through the cloud platform based on cloud computing, the supply chain data sharing is realized. Tag moving scenarios usually use a single reader. The proposed employs double reader to expand the recognition range and also to prolong the recognition time, but there is the problem of repeated recognition. This article proposes a suffix-based bit-slot ALOHA (SBSA) algorithm, where this problem is solved and the time-slot is reduced. In view of the problem that a large number of tags entering into the recognition area of readers lead to the reduction of the recognition efficiency. This article proposes a suffix-based tag group bit-slot ALOHA (STGBSA) algorithm, tag grouping and tag estimation methods are proposed. Complexity ...
Online Analytical Processing provides multidimensional, summarized views of business data. It ena... more Online Analytical Processing provides multidimensional, summarized views of business data. It enables users to drill down or roll up from a level to another level in a dimension, which forms an analytical path/thread. In many scenarios, this analytical path is very useful for users to keep their analysis in context. But most visualization tools display only the result in one chart after drilling down or rolling up. The analytical context or the intermediate process is lost or indirect. This paper presents a novel OLAP visualization approach called TAVO - a Tree-like Analysis View for OLAP. It visualizes the multi-dimensional data by various charts and organizes them in a tree, where each node is a chart and connects other charts when a user navigates along a dimension by levels or to another dimension. By this way, the drill-down/roll-up semantic relationship is clearly expressed. Users can easily explore relative history queries by rolling up in a chart. We implemented TAVO by dyna...
The data cube is a critical tool for accelerating online analysis in big data. Due to its exponen... more The data cube is a critical tool for accelerating online analysis in big data. Due to its exponential space overhead, the quotient cube, as the main data cube compression approach, was proposed to significantly reduce the number of data cells if they are aggregated over the same base tuple set, i.e. they are cover equivalent to form an equivalence class. Nevertheless, it still poses challenges to efficiently analyze massive data due to high storage space consumption. This paper proposes the reduced quotient cube (RQC) based on the following observation. (i) there are equivalence classes of various sizes in a quotient cube; (ii) the small equivalence classes usually dominate; (iii) the big equivalence classes are more capable of query answering since they can induce more data cells. Unlike the quotient cube, which preserves all the equivalence classes of equal priority, the reduced quotient cube preferentially does those with larger query answering capacity and smaller space occupied capacity. Further, we design its efficient constructing and querying algorithms. The extensive experimental results show that compared with the quotient cube, the reduced quotient cube space is only 11.3% while the maximum query capacity is 95.9%. The query time of the reduced quotient cube is reduced by 51.24% on average compared to the quotient cube.
2021 IEEE 3rd International Conference on Computer Science and Educational Informatization (CSEI)
With the continuous development of society and the rapid growth of data volume, the impact of big... more With the continuous development of society and the rapid growth of data volume, the impact of big data analysis has become increasingly prominent on social decision-making, and the application of big data analysis to education decision-making has also become an important issue. This article takes the third-grade non-major course performance of a primary school in a certain city as the research object, analyzes the influence of teacher’s teaching on student performance based on machine language learning and statistical methods. The results show that gender, education background, professional title, and attitude have a significantly positive impact on students’ English, etc., non-major course performance.
2021 IEEE 3rd International Conference on Computer Science and Educational Informatization (CSEI)
College entrance examination and general education play important roles in university study. Yet,... more College entrance examination and general education play important roles in university study. Yet, the mechanism of how they impact the students’ graduation achievement is still not clear. This paper uses the students’ test scores enrolled in a university in Yunnan Province, China, from 2012 to 2015 as the sample object. By establishing a multiple linear regression model, the paper explores the impact of college entrance examination scores and general education scores on graduation scores. The linear regression model established in this study shows that college entrance examinations and public courses have a significant impact on graduation achievements. Finally, through the case study of 13,607 sample data, the predicted results are basically the same as the actual results, which further illustrates this research’s reliability and accuracy.
Superset query is widely used in object-oriented databases, data mining, and many other fields. T... more Superset query is widely used in object-oriented databases, data mining, and many other fields. Trie is an efficient index for superset query, whereas most existing trie index aim at improving query performance while ignoring storage overheads. To solve this problem, in this paper, we propose an efficient extended Level-Ordered Unary Degree Sequence (LOUDS) index: Ext-LOUDS. Ext-LOUDS expresses a trie by 1 integer vector and 3 bit vectors directly map each NodeID to its corresponding position, thus accelerating some key operations needed for superset query. Based on Ext-LOUDS, an efficient superset query algorithm, ELOUDS-Super, is designed. Experimental results on both real and synthetic datasets show that Ext-LOUDS can decrease 50%–60% space overheads compared with trie while maintaining a relative good query performance.
There are two important models for data analysis and knowledge system: data cube lattices and con... more There are two important models for data analysis and knowledge system: data cube lattices and concept lattices. They both essentially have lattice structures, which are actually irregular in our real world. However, their structural characteristics and relationship are not yet clear. To the best of our knowledge, no work has paid enough attention to this challenging issue from the perspective of graph data, in spite of the importance of structures in lattice data. In this paper, we first tackle the structural statistics of lattice data from three aspects: the degree distribution, clustering coefficient, and average path length. We demonstrated by various datasets that data cube lattices and concept lattices share similarities underlying their topology, which are, in general, different from random networks and complex networks. Specifically, lattice data follow the Poisson distribution and have smaller clustering coefficient and greater average path length. We further discuss and exp...
ABSTRACT Due to current real-time data compression algorithms is not efficient enough, we have pr... more ABSTRACT Due to current real-time data compression algorithms is not efficient enough, we have proposed a two-phase real-time data compression algorithm which can be very fast in data compression with high compression rate. The algorithm can adapt to both text and binary files. The first phase compresses the file with long common strings into short forms. The second phase compresses the result of the first phase with short bytes in common into the final results. We name the algorithm TDSC Algorithm.
Communications in Computer and Information Science, 2013
To obtain the insight in a single large graph and to save the space consumption for graph mining,... more To obtain the insight in a single large graph and to save the space consumption for graph mining, the graph summary transforms the input graph into an aggregated concise super-graph represented by supernodes and superedges. In this paper, we investigate current algorithms of the graph summary and aggregation. We provide the classification of them in terms of partition criterion or information lossless. Further, the main graph summary algorithms are compared and discussed in detail. In the end, we give the challenges and future works.
ABSTRACT In order to address the visualized analysis on the multi-dimensional data, this demo pap... more ABSTRACT In order to address the visualized analysis on the multi-dimensional data, this demo paper designs and implements a visualization analysis tool, named as Bizard. Bizard encapsulates the underlying Hadoop Client or XML/A data access interface, which makes it convenient to use conventional analytical tools. By Rich Internet Application (RIA) technology, Bizard designs multiple visualization dashboard widgets and supports interactive multi-dimensional analysis.
Graph representation learning aims to represent the structural and semantic information of graph ... more Graph representation learning aims to represent the structural and semantic information of graph objects as dense real value vectors in low dimensional space by machine learning. It is widely used in node classification, link prediction, and recommendation systems. However, directly computing the embeddings for original graphs is prohibitively inefficient, especially for large-scale graphs. To address this issue, we present the GSE (Graph Summarization Embedding) model, a more efficient model that computes the nodes' embeddings based on graph summarization. Specifically, the model first searches for the minimum information entropy of k groups to transform the original graph into a hypergraph with higherorder structural features. Next, the summarization graph's connection probabilities are used to determine the biased random walks on the hypergraph, which then generates the sequences of the super-nodes. Finally, the node sequences are fed into the skip-gram to generate the vectors of these nodes. Our proposed model improves the efficiency of graph embedding on big data graphs and effectively alleviates the local optimal problem caused by the random walks. Experimental results demonstrate that GSE outperforms main existing clustering baselines, such as K_Means Clustering, Affinity Propagation Clustering, Canopy Clustering, and ACP Clustering. Moreover, our model can be coupled with the main graph embedding methods and improves the Macro-F1 scores and Micro-F1 scores for classification tasks on a variety of real-world graph data. INDEX TERMS Graph embedding, graph summarization, graph coarsening, random walks.
Closed cubing is a very efficient algorithm for data cube
compression proposed recently in the l... more Closed cubing is a very efficient algorithm for data cube
compression proposed recently in the literature. It
losslessly condenses a group of cells into one cell if these
cells have the same aggregate value and preserve rollup/
drill-down semantics. Despite its importance, parallel
closed cubing solutions for huge data sets are not well
studied so far to the best of the authors’ knowledge. This
paper presents a parallel closed cube construction and
query algorithm over low cost PC clusters using the
MapReduce framework. In addition, we proved that with
the number of data blocks increases, the closed cubes’
storage size decreases gradually. Thus users can specify the
number of data blocks to balance the performance between
cubes storage and query time. Experimental study
demonstrates that our algorithm is efficient and scalable.
With rapid development of the Internet, the original knowledge management systems which are centr... more With rapid development of the Internet, the original knowledge management systems which are centralized control can not be adaptive to the distributed environment, because there maybe more than one knowledge bases including personal knowledge base in the collaborative ...
Incomplete information about how data has evolved over time has caused many puzzles in applicatio... more Incomplete information about how data has evolved over time has caused many puzzles in applications such as audit trail, data quality. Such challenges mandate the need for rich and descriptive derivation information about data to accompany the data in order to understand and reuse it. This ...
INTRODUCTION: The index is one of the most effective ways to improve the database query performan... more INTRODUCTION: The index is one of the most effective ways to improve the database query performance. The expert-based index recommendation approach cannot adjust the index configuration in real time. At the same time, reinforcement learning can automatically update the index and improve the recommended configuration by leveraging expert experience.OBJECTIVES: This paper proposes the RBOIRA, which combines rules and reinforcement learning to recommend the optimal index configuration for a set of workloads in a dynamic database.METHODS: Firstly, RBOIRA designed three heuristic rules for pruning index candidates. Secondly, it uses reinforcement learning to recommend the optimal index configuration for a set of workloads in the database. Finally, we conducted extensive experiments to evaluate RBOIRA using the TPC-H database benchmark.RESULTS: RBOIRA recommends index configurations with superior performance compared to the baselines we define and other reinforcement learning methods used...
Recent years have seen a surge in approaches that combine deep learning and recommendation system... more Recent years have seen a surge in approaches that combine deep learning and recommendation systems to capture user preference or item interaction evolution over time. However, the most related work only consider the sequential similarity between the items and neglects the item content feature information and the impact difference of interacted items on the next items. This paper introduces the deep bidirectional long short-term memory (LSTM) and self-attention mechanism into the sequential recommender while fusing the information of item sequences and contents. Specifically, we deal with the issues in a three-pronged attack: the improved item embedding, weight update, and the deep bidirectional LSTM preference learning. First, the user-item sequences are embedded into a low-dimensional item vector space representation via Item2vec, and the class label vectors are concatenated for each embedded item vector. Second, the embedded item vectors learn different impact weights of each item...
Airfare price prediction is one of the core facilities of the decision support system in civil av... more Airfare price prediction is one of the core facilities of the decision support system in civil aviation, which includes departure time, days of purchase in advance and flight airline. The traditional airfare price prediction system is limited by the nonlinear interrelationship of multiple factors and fails to deal with the impact of different time steps, resulting in low prediction accuracy. To address these challenges, this paper proposes a novel civil airline fare prediction system with a Multi-Attribute Dual-stage Attention (MADA) mechanism integrating different types of data extracted from the same dimension. In this method, the Seq2Seq model is used to add attention mechanisms to both the encoder and the decoder. The encoder attention mechanism extracts multi-attribute data from time series, which are optimized and filtered by the temporal attention mechanism in the decoder to capture the complex time dependence of the ticket price sequence. Extensive experiments with actual ci...
At present, some existing tag anti-collision algorithms are used mostly in scenarios of fixed num... more At present, some existing tag anti-collision algorithms are used mostly in scenarios of fixed number of tags and they perform poorly in tag moving scenarios. Therefore, this paper proposes a double reader tag movement RFID anti-collision method based on suffix, and through the cloud platform based on cloud computing, the supply chain data sharing is realized. Tag moving scenarios usually use a single reader. The proposed employs double reader to expand the recognition range and also to prolong the recognition time, but there is the problem of repeated recognition. This article proposes a suffix-based bit-slot ALOHA (SBSA) algorithm, where this problem is solved and the time-slot is reduced. In view of the problem that a large number of tags entering into the recognition area of readers lead to the reduction of the recognition efficiency. This article proposes a suffix-based tag group bit-slot ALOHA (STGBSA) algorithm, tag grouping and tag estimation methods are proposed. Complexity ...
Online Analytical Processing provides multidimensional, summarized views of business data. It ena... more Online Analytical Processing provides multidimensional, summarized views of business data. It enables users to drill down or roll up from a level to another level in a dimension, which forms an analytical path/thread. In many scenarios, this analytical path is very useful for users to keep their analysis in context. But most visualization tools display only the result in one chart after drilling down or rolling up. The analytical context or the intermediate process is lost or indirect. This paper presents a novel OLAP visualization approach called TAVO - a Tree-like Analysis View for OLAP. It visualizes the multi-dimensional data by various charts and organizes them in a tree, where each node is a chart and connects other charts when a user navigates along a dimension by levels or to another dimension. By this way, the drill-down/roll-up semantic relationship is clearly expressed. Users can easily explore relative history queries by rolling up in a chart. We implemented TAVO by dyna...
The data cube is a critical tool for accelerating online analysis in big data. Due to its exponen... more The data cube is a critical tool for accelerating online analysis in big data. Due to its exponential space overhead, the quotient cube, as the main data cube compression approach, was proposed to significantly reduce the number of data cells if they are aggregated over the same base tuple set, i.e. they are cover equivalent to form an equivalence class. Nevertheless, it still poses challenges to efficiently analyze massive data due to high storage space consumption. This paper proposes the reduced quotient cube (RQC) based on the following observation. (i) there are equivalence classes of various sizes in a quotient cube; (ii) the small equivalence classes usually dominate; (iii) the big equivalence classes are more capable of query answering since they can induce more data cells. Unlike the quotient cube, which preserves all the equivalence classes of equal priority, the reduced quotient cube preferentially does those with larger query answering capacity and smaller space occupied capacity. Further, we design its efficient constructing and querying algorithms. The extensive experimental results show that compared with the quotient cube, the reduced quotient cube space is only 11.3% while the maximum query capacity is 95.9%. The query time of the reduced quotient cube is reduced by 51.24% on average compared to the quotient cube.
2021 IEEE 3rd International Conference on Computer Science and Educational Informatization (CSEI)
With the continuous development of society and the rapid growth of data volume, the impact of big... more With the continuous development of society and the rapid growth of data volume, the impact of big data analysis has become increasingly prominent on social decision-making, and the application of big data analysis to education decision-making has also become an important issue. This article takes the third-grade non-major course performance of a primary school in a certain city as the research object, analyzes the influence of teacher’s teaching on student performance based on machine language learning and statistical methods. The results show that gender, education background, professional title, and attitude have a significantly positive impact on students’ English, etc., non-major course performance.
2021 IEEE 3rd International Conference on Computer Science and Educational Informatization (CSEI)
College entrance examination and general education play important roles in university study. Yet,... more College entrance examination and general education play important roles in university study. Yet, the mechanism of how they impact the students’ graduation achievement is still not clear. This paper uses the students’ test scores enrolled in a university in Yunnan Province, China, from 2012 to 2015 as the sample object. By establishing a multiple linear regression model, the paper explores the impact of college entrance examination scores and general education scores on graduation scores. The linear regression model established in this study shows that college entrance examinations and public courses have a significant impact on graduation achievements. Finally, through the case study of 13,607 sample data, the predicted results are basically the same as the actual results, which further illustrates this research’s reliability and accuracy.
Superset query is widely used in object-oriented databases, data mining, and many other fields. T... more Superset query is widely used in object-oriented databases, data mining, and many other fields. Trie is an efficient index for superset query, whereas most existing trie index aim at improving query performance while ignoring storage overheads. To solve this problem, in this paper, we propose an efficient extended Level-Ordered Unary Degree Sequence (LOUDS) index: Ext-LOUDS. Ext-LOUDS expresses a trie by 1 integer vector and 3 bit vectors directly map each NodeID to its corresponding position, thus accelerating some key operations needed for superset query. Based on Ext-LOUDS, an efficient superset query algorithm, ELOUDS-Super, is designed. Experimental results on both real and synthetic datasets show that Ext-LOUDS can decrease 50%–60% space overheads compared with trie while maintaining a relative good query performance.
There are two important models for data analysis and knowledge system: data cube lattices and con... more There are two important models for data analysis and knowledge system: data cube lattices and concept lattices. They both essentially have lattice structures, which are actually irregular in our real world. However, their structural characteristics and relationship are not yet clear. To the best of our knowledge, no work has paid enough attention to this challenging issue from the perspective of graph data, in spite of the importance of structures in lattice data. In this paper, we first tackle the structural statistics of lattice data from three aspects: the degree distribution, clustering coefficient, and average path length. We demonstrated by various datasets that data cube lattices and concept lattices share similarities underlying their topology, which are, in general, different from random networks and complex networks. Specifically, lattice data follow the Poisson distribution and have smaller clustering coefficient and greater average path length. We further discuss and exp...
ABSTRACT Due to current real-time data compression algorithms is not efficient enough, we have pr... more ABSTRACT Due to current real-time data compression algorithms is not efficient enough, we have proposed a two-phase real-time data compression algorithm which can be very fast in data compression with high compression rate. The algorithm can adapt to both text and binary files. The first phase compresses the file with long common strings into short forms. The second phase compresses the result of the first phase with short bytes in common into the final results. We name the algorithm TDSC Algorithm.
Communications in Computer and Information Science, 2013
To obtain the insight in a single large graph and to save the space consumption for graph mining,... more To obtain the insight in a single large graph and to save the space consumption for graph mining, the graph summary transforms the input graph into an aggregated concise super-graph represented by supernodes and superedges. In this paper, we investigate current algorithms of the graph summary and aggregation. We provide the classification of them in terms of partition criterion or information lossless. Further, the main graph summary algorithms are compared and discussed in detail. In the end, we give the challenges and future works.
ABSTRACT In order to address the visualized analysis on the multi-dimensional data, this demo pap... more ABSTRACT In order to address the visualized analysis on the multi-dimensional data, this demo paper designs and implements a visualization analysis tool, named as Bizard. Bizard encapsulates the underlying Hadoop Client or XML/A data access interface, which makes it convenient to use conventional analytical tools. By Rich Internet Application (RIA) technology, Bizard designs multiple visualization dashboard widgets and supports interactive multi-dimensional analysis.
Graph representation learning aims to represent the structural and semantic information of graph ... more Graph representation learning aims to represent the structural and semantic information of graph objects as dense real value vectors in low dimensional space by machine learning. It is widely used in node classification, link prediction, and recommendation systems. However, directly computing the embeddings for original graphs is prohibitively inefficient, especially for large-scale graphs. To address this issue, we present the GSE (Graph Summarization Embedding) model, a more efficient model that computes the nodes' embeddings based on graph summarization. Specifically, the model first searches for the minimum information entropy of k groups to transform the original graph into a hypergraph with higherorder structural features. Next, the summarization graph's connection probabilities are used to determine the biased random walks on the hypergraph, which then generates the sequences of the super-nodes. Finally, the node sequences are fed into the skip-gram to generate the vectors of these nodes. Our proposed model improves the efficiency of graph embedding on big data graphs and effectively alleviates the local optimal problem caused by the random walks. Experimental results demonstrate that GSE outperforms main existing clustering baselines, such as K_Means Clustering, Affinity Propagation Clustering, Canopy Clustering, and ACP Clustering. Moreover, our model can be coupled with the main graph embedding methods and improves the Macro-F1 scores and Micro-F1 scores for classification tasks on a variety of real-world graph data. INDEX TERMS Graph embedding, graph summarization, graph coarsening, random walks.
Closed cubing is a very efficient algorithm for data cube
compression proposed recently in the l... more Closed cubing is a very efficient algorithm for data cube
compression proposed recently in the literature. It
losslessly condenses a group of cells into one cell if these
cells have the same aggregate value and preserve rollup/
drill-down semantics. Despite its importance, parallel
closed cubing solutions for huge data sets are not well
studied so far to the best of the authors’ knowledge. This
paper presents a parallel closed cube construction and
query algorithm over low cost PC clusters using the
MapReduce framework. In addition, we proved that with
the number of data blocks increases, the closed cubes’
storage size decreases gradually. Thus users can specify the
number of data blocks to balance the performance between
cubes storage and query time. Experimental study
demonstrates that our algorithm is efficient and scalable.
With rapid development of the Internet, the original knowledge management systems which are centr... more With rapid development of the Internet, the original knowledge management systems which are centralized control can not be adaptive to the distributed environment, because there maybe more than one knowledge bases including personal knowledge base in the collaborative ...
Incomplete information about how data has evolved over time has caused many puzzles in applicatio... more Incomplete information about how data has evolved over time has caused many puzzles in applications such as audit trail, data quality. Such challenges mandate the need for rich and descriptive derivation information about data to accompany the data in order to understand and reuse it. This ...
Uploads
Papers by Jinguo You
compression proposed recently in the literature. It
losslessly condenses a group of cells into one cell if these
cells have the same aggregate value and preserve rollup/
drill-down semantics. Despite its importance, parallel
closed cubing solutions for huge data sets are not well
studied so far to the best of the authors’ knowledge. This
paper presents a parallel closed cube construction and
query algorithm over low cost PC clusters using the
MapReduce framework. In addition, we proved that with
the number of data blocks increases, the closed cubes’
storage size decreases gradually. Thus users can specify the
number of data blocks to balance the performance between
cubes storage and query time. Experimental study
demonstrates that our algorithm is efficient and scalable.
compression proposed recently in the literature. It
losslessly condenses a group of cells into one cell if these
cells have the same aggregate value and preserve rollup/
drill-down semantics. Despite its importance, parallel
closed cubing solutions for huge data sets are not well
studied so far to the best of the authors’ knowledge. This
paper presents a parallel closed cube construction and
query algorithm over low cost PC clusters using the
MapReduce framework. In addition, we proved that with
the number of data blocks increases, the closed cubes’
storage size decreases gradually. Thus users can specify the
number of data blocks to balance the performance between
cubes storage and query time. Experimental study
demonstrates that our algorithm is efficient and scalable.