Filters








294,002 Hits in 4.0 sec

An Analysis of Virtual Loss in Parallel MCTS

S. Ali Mirsoleimani, Aske Plaat, Jaap van den Herik, Jos Vermaseren
2017 Proceedings of the 9th International Conference on Agents and Artificial Intelligence  
Monte Carlo tree search algorithms, such as UCT, select the best-root-child as a result of an iterative search process consistent with path dependency.  ...  The results show that using virtual loss for lock-free tree parallelization still degrades the performance of the algorithm. Contrary to what we aimed at.  ...  Therefore, a method called virtual loss is proposed for a lock-based tree parallelizations to force parallel workers to traverse different paths inside the MCTS tree.  ... 
doi:10.5220/0006205806480652 dblp:conf/icaart/MirsoleimaniPHV17 fatcat:iu3oo665l5b75kumdiei3ndayy

Asynch-SGBDT: Asynchronous Parallel Stochastic Gradient Boosting Decision Tree based on Parameters Server [article]

Cheng Daning, Xia Fen, Li Shigang, Zhang Yunquan
2019 arXiv   pre-print
Experimental results also show asynch-SGBDT training process reaches a linear speedup in asynchronous parallel manner when datasets and GBDT trees meet high scalability requirements.  ...  However, those GBDT algorithms are synchronous parallel algorithms which fail to make full use of Parameter Server.  ...  Parallel GBDT Training Algorithm Although many studies offered different work about GBDT, current GBDT algorithm uses fork-join parallel method to parallelize GBDT training process.  ... 
arXiv:1804.04659v4 fatcat:qddpqdarszhexbufa7gxodjbuu

Parallel univariate decision trees

Olcay Taner Yıldız, Onur Dikmen
2007 Pattern Recognition Letters  
In this paper, we first present two different univariate decision tree algorithms C4.5 and univariate linear discriminant tree.  ...  In those cases it is highly desirable to construct univariate decision trees in reasonable time. This may be accomplished by parallelizing univariate decision tree algorithms.  ...  Acknowledgement We thank to Burak Gü rdag for fruitful discussions on parallel computing and his help in using Asma.  ... 
doi:10.1016/j.patrec.2006.11.009 fatcat:be6vsf6rnvgcxcmbmqscp5l4oq

Load Balancing and Parallel Multiple Sequence Alignment with Tree Accumulation [chapter]

Guangming Tan, Liu Peng, Shengzhong Feng, Ninghui Sun
2006 Lecture Notes in Computer Science  
We proposed a load balancing strategy for parallelizing tree accumulation in progressive alignment.  ...  The new parallel progressive alignment algorithm reducing to tree accumulation with load balancing reduced the overall running time greatly and achieved reasonable speedups.  ...  Table 1 demonstrates the overall running time for two parallel algorithms with the different number of processors and different size of problems.  ... 
doi:10.1007/11823285_120 fatcat:sgdmypjbzbdizdl257ldrn6uau

Alphabet-dependent Parallel Algorithm for Suffix Tree Construction for Pattern Searching [article]

Freeson Kaniwa, Venu Madhav Kuthadi, Otlhapile Dinakenyane, Heiko Schroeder
2017 arXiv   pre-print
We present a novel alphabet-dependent parallel algorithm which attempts to take advantage of the perverseness of the multicore architecture.  ...  We experimentally achieved up to 15x speedup over the sequential algorithm on different input sizes of biological sequences.  ...  suffix tree) whereas ours implements a parallel algorithm of k-mer based of a sequential algorithm for the generalized suffix tree.  ... 
arXiv:1704.05660v1 fatcat:55uj7sdchjet5kcibtywixnxai

Alphabet-dependent Parallel Algorithm for Suffix Tree Construction for Pattern Searching

Freeson Kaniwa, Venu Madhav Kuthadi, Otlhapile Dinakenyane, Heiko Schroeder
2017 International Journal of Grid and Distributed Computing  
We present a novel alphabet-dependent parallel algorithm which attempts to take advantage of the perverseness of the multicore architecture.  ...  We experimentally achieved up to 15x speedup over the sequential algorithm on different input sizes of biological sequences.  ...  suffix tree) whereas ours implements a parallel algorithm of k-mer based of a sequential algorithm for the generalized suffix tree.  ... 
doi:10.14257/ijgdc.2017.10.1.02 fatcat:6nfccy5hezbjdjd2spgmjrfo6e

Techniques for Designing Efficient Parallel Graph Algorithms for SMPs and Multicore Processors [chapter]

Guojing Cong, David A. Bader
2007 Lecture Notes in Computer Science  
We propose techniques for designing and implementing efficient parallel algorithms for graph problems on symmetric multiprocessors and chip multiprocessors with a case study of parallel tree and connectivity  ...  The problems we study represent a wide range of irregular problems that have fast theoretic parallel algorithms but no known efficient parallel implementations that achieve speedup without serious restricting  ...  Note that different runs of the algorithm may produce trees of different topologies, yet each is a correct spanning tree.  ... 
doi:10.1007/978-3-540-74742-0_15 fatcat:qmtryl6qmnfavk45kbxls5bzca

A Dynamic Approach to Weighted Suffix Tree Construction Algorithm

Binay Kumar Pandey, Rajdeep Niyogi, Ankush Mittal
2011 International Journal of Distributed and Parallel systems  
Although a static partitioning based parallel algorithm existed for the construction of weighted suffix tree, but for very long weighted DNA sequences it takes significant amount of time.  ...  However, in our implementation of dynamic partition based parallel weighted suffix tree construction algorithm on cluster computing makes it possible to significantly accelerate the construction of weighted  ...  Implementation of static weighted suffix tree construction algorithm on Cluster computing using MPI The parallelization of weighted suffix tree construction algorithm on Cluster computing is a bit different  ... 
doi:10.5121/ijdps.2011.2103 fatcat:vz4h4z3rnzd45gjjq2rsw2ykmy

A New Parallel Matrix Multiplication Algorithm on Tree-Hypercube Network using Iman1 Supercomputer

Orieb AbuAlghanam, Mohammad Qatawneh, Hussein A., Omar Adwan, Ammar Huneiti
2017 International Journal of Advanced Computer Science and Applications  
The tree-hypercube (TH) interconnection network is relatively a new interconnection network, which is constructed from tree and hypercube topologies.  ...  Fig. 9 displays the parallel efficiency for matrix multiplication according to different number of processors and different matrix size.  ...  Pseudocode Tree-Hypercube Algorithm for Matrix Multiplication. www.ijacsa.thesai.org Fig. 4 . 4 Parallel matrix multiplication tree-hypercube view. Fig.  ... 
doi:10.14569/ijacsa.2017.081225 fatcat:dxbz4ggi4rfbbltypysthtwqwa

Parallelization of a Hierarchical Graph-Based Image Segmentation using OpenMP

Ali Sağlam, Nurdan Akhan Baykan
2016 International Journal of Applied Mathematics Electronics and Computers  
For this reason, the algorithm to be parallelized must be suitable to parallel.  ...  In this paper, we provide a parallel implementation to a hierarchical graph-based image segmentation method by using its hierarchical processing steps.  ...  An internal difference of a tree is the weight value of the highest weighted edge in the tree. 1 and 2 defines the number of vertices in the trees 1 and 2 respectively.  ... 
doi:10.18100/ijamec.271038 fatcat:iusn64psqzfovpratpr5gcxl2y

Methods and Evaluations of Decision Tree Algorithms on GPUs: An Overview

Nesreen Adnan Hamad, Fatima Mousa Quiam, Khalid Mohammad Jaber
2018 ICIC Express Letters  
a decision tree.  ...  However, when dealing with massive datasets, the time needed to build the decision tree will increase. Therefore, parallel computing is used to accelerate the construction of the decision tree.  ...  Different Recent Parallel Techniques for Decision Tree. This section discusses the decision tree using different recent parallel techniques other than the GPU technique.  ... 
doi:10.24507/icicel.12.07.723 fatcat:y7bt7qub75dbpnlkwzlrr3kjdy

Improved Balanced Parallel FP-Growth with MapReduce

Qing YANG, Fei-Yang DU, Xi ZHU, Cheng-Gong JIANG
2017 DEStech Transactions on Computer Science and Engineering  
By using Cutting strategy on PFP which is the original algorithm with MapReduce, we merged paths which are not frequent in FP-tree, and designed a new parallel FP-Growth algorithm.  ...  Through the combination of these two strategies, this paper designed a new parallel FP-Growth algorithm.  ...  The FP-Growth Algorithm Based On Cutting Strategy We designed a frequent item sets mining algorithm based on cutting strategy, Cutting FP-tree algorithm.  ... 
doi:10.12783/dtcse/aice-ncs2016/5681 fatcat:2i3pro6wqbe4rh37dir6hf63pq

Parallel query execution in PRISMA/DB [chapter]

Annita N. Wilschut, Peter M. G. Apers, Jan Flokstra
1991 Lecture Notes in Computer Science  
A n n i t a N. W i l s c h u t P e t e r M. G. A p e r s J a n F l o k s t r a U n i v e r s i t y of T w e n t e  ...  in a query tree is distributed over more than one processor. i n t e r -o p e r a t o r p a r a l l e l i s m Different operations in one query tree are executed concurrently. i n t e r -q u e r y parallelism  ...  A pipelining hash-join algorithm To increase the amount of effective parallelism in the execution of the join-tree in figure 2, a new main-memory hash-join algorithm, called pipelining hash-join is proposed  ... 
doi:10.1007/3-540-54132-2_69 fatcat:uczylg5dazeptchytxerwve5am

Regularity Versus Load-balancing on GPU for Treefix Computations

David Defour, Manuel Marin
2013 Procedia Computer Science  
This article discusses of both solutions for treefix operations, which consist of applying a certain operation while performing a tree traversal.  ...  The use of GPUs has enabled us to achieve substantial acceleration in highly regular data parallel applications.  ...  For other cases, when there is no bound on the number of children, nor on the tree topology, a different algorithm has to be used.  ... 
doi:10.1016/j.procs.2013.05.194 fatcat:wnozcqn5vzcrhdrdklu6ystq2y

Parallel UCT search on GPUs

Nicolas A. Barriga, Marius Stanescu, Michael Buro
2014 2014 IEEE Conference on Computational Intelligence and Games  
We propose two parallel UCT search (Upper Confidence bounds applied to Trees) algorithms that take advantage of modern GPU hardware.  ...  Empirical results show that our proposed Multiblock Parallel algorithm outperforms other approaches and can take advantage of the GPU hardware without the added complexity of searching multiple trees.  ...  Figure 13 compares Block and Multiblock algorithms using this setting. For the Block Parallel algorithm the number of trees didn't make a significant difference.  ... 
doi:10.1109/cig.2014.6932879 dblp:conf/cig/BarrigaSB14 fatcat:aw3gvsnvmncohm2qxgnfyif5na
« Previous Showing results 1 — 15 out of 294,002 results