GPU Accelerated Computation of the Longest Common Subsequence

Kawanami, Katsuya; Fujimoto, Noriyuki

doi:10.1007/978-3-642-30397-5_8

Cited by 6 publications

(4 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We explored different methods and ways to parallelize our code, in order to deploy it for human robot interaction in real time. We used the GPU-LCS algorithm in [14] for the LCS-based classification process that is explained in Sect. 3.5.…”

Section: Parallelizing the Classification Algorithmmentioning

confidence: 99%

Using patterns of firing neurons in spiking neural networks for learning and early recognition of spatio-temporal patterns

Rekabdar

Nicolescu

Louis

2016

Neural Comput & Applic

View full text Add to dashboard Cite

In this paper, we propose a novel unsupervised learning approach for spatio-temporal pattern classification. We use a spike timing neural network with axonal conductance delays to learn the structure of spatio-temporal patterns from a small set of training samples and then use this network for classifying unseen patterns. The method also enables early classification of patterns, before they are completely observed. To transform a spatio-temporal pattern into a suitable form for the spiking network, we create a mapping process that transforms it into a spike train that can be used to train the network through spiketiming-dependent plasticity. Based on the trained network, we build models of the training samples as strings of characters in which each character represents a set of neurons that fire at a particular time step, in response to the pattern. For classification, we compute the longest common subsequence between the model strings corresponding to the input and training samples and choose the class of the training sample with highest similarity. This method is evaluated on a handwritten digits dataset with spatio-temporal information. We show that this method is robustly detecting the correct class early on. Comparison with one unsupervised and eight other supervised approaches shows that our proposed approach could compete or has better performance rather than all of them. Further analysis show that even for misclassified samples, our system can detect the correct class among the top three class labels.

show abstract

Section: Parallelizing the Classification Algorithmmentioning

confidence: 99%

Using patterns of firing neurons in spiking neural networks for learning and early recognition of spatio-temporal patterns

Rekabdar

Nicolescu

Louis

2016

Neural Comput & Applic

View full text Add to dashboard Cite

show abstract

“…The GPU adaptation of bit-vector solutions is very rare. One recent work that uses both bit-wise operations and GPUs is by Kawanami and Fujimoto (2012) who give a GPU implementation of the one-to-one longest common subsequence (LCS) algorithm. Ozsoy et al (2013Ozsoy et al ( , 2014a investigated one-to-many LCS and scoring on GPUs using bit-vector approach.…”

Section: Related Workmentioning

confidence: 99%

“…It may require deep investigation of the problem and redesigning the traditional algorithms for GPU capabilities. One such approach that shows great benefit from GPGPUs is the bit-vector algorithms on string matching domain (Kawanami and Fujimoto, 2012;Ozsoy et al, 2013Ozsoy et al, , 2014a. Bit-vector parallelism provides intrinsic parallelism within a computer word, where each bit in a word represents a data point that uses bit-wise operations to update all the positions in the same word at the same time.…”

Section: Introductionmentioning

confidence: 99%

An efficient parallelization of longest prefix match and application on data compression

Özsoy

2016

The International Journal of High Performance Computing Applica

View full text Add to dashboard Cite

In this article, we describe a new approach to parallelize longest prefix match (LPM) algorithm through bit parallelism, also known as bit-vector approach. This approach makes use of bit-wise computations and leverages bit parallelism. The proposed parallel algorithm will be demonstrated in dictionary-based lossless data compression on general-purpose graphics processing units (GPGPUs). One of the main contributions of this work is redesigning the core part of the data compression algorithm and replacing it with the newly proposed bit-vector LPM solution. Using bit parallelism is a fundamentally new approach for data compression and promising in performance for hybrid CPU-GPU environments. The implementation of the new compression algorithm on GPUs improves the performance of the compression process compared to the previous attempts. Moreover, the bit-vector approach opens new opportunities for improvement and increases the applicability to popular heterogeneous environments.

show abstract

“…Bit-level parallelism can also be exploited for the Longest Common Subsequence (LCS) problem, which is similar to, but simpler than, the problem of computing Levenshtein distance. Kawanami and Fujimoto [8] implement the first GPU solution, which exploits task parallelism. Ozsoy et al [21] propose an improved GPU design that reaches 1 TCUPS using 3 Fermi GPUs.…”

Section: Related Workmentioning

confidence: 99%

Thread-cooperative, bit-parallel computation of levenshtein distance on GPU

Chacón

Marco-Sola

Espinosa

et al. 2014

Proceedings of the 28th ACM International Conference on Supercomputing

View full text Add to dashboard Cite

Approximate string matching is a very important problem in computational biology; it requires the fast computation of string distance as one of its essential components. Myers' bit-parallel algorithm improves the classical dynamic programming approach to Levenshtein distance computation, and offers competitive performance on CPUs. The main challenge when designing an efficient GPU implementation is to expose enough SIMD parallelism while at the same time keeping a relatively small working set for each thread. In this work we implement and optimise a CUDA version of Myers' algorithm suitable to be used as a building block for DNA sequence alignment. We achieve high efficiency by means of a cooperative parallelisation strategy for (1) very-long integer addition and shift operations, and (2) several simultaneous pattern matching tasks. In addition, we explore the performance impact obtained when using features specific to the Kepler architecture. Our results show an overall performance of the order of tera cells updates per second using a single high-end Nvidia GPU, and factor speedups in excess of 20× with respect to a sixteen-core, non-vectorised CPU implementation.

show abstract

GPU Accelerated Computation of the Longest Common Subsequence

Cited by 6 publications

References 6 publications

Using patterns of firing neurons in spiking neural networks for learning and early recognition of spatio-temporal patterns

Using patterns of firing neurons in spiking neural networks for learning and early recognition of spatio-temporal patterns

An efficient parallelization of longest prefix match and application on data compression

Thread-cooperative, bit-parallel computation of levenshtein distance on GPU

Contact Info

Product

Resources

About