Re-ranking algorithms have been proposed to improve the effectiveness of content-based image retrieval systems by exploiting contextual information encoded in distance measures and ranked lists. In this paper, we show...
详细信息
Re-ranking algorithms have been proposed to improve the effectiveness of content-based image retrieval systems by exploiting contextual information encoded in distance measures and ranked lists. In this paper, we show how we improved the efficiency of one of these algorithms, called Contextual Spaces Re-Ranking (CSRR). One of our approaches consists in parallelizing the algorithm with OpenCL to use the central and graphics processing units of an accelerated processing unit. The other is to modify the algorithm to a version that, when compared with the original CSRR, not only reduces the total running time of our implementations by a median of 1.6x but also increases the accuracy score in most of our test cases. Combining both parallelization and algorithm modification results in a median speedup of 5.4x from the original serial CSRR to the parallelized modified version. Different implementations for CSRR's Re-sort Ranked Lists step were explored as well, providing insights into graphics processing unit sorting, the performance impact of image descriptors, and the trade-offs between effectiveness and efficiency. Copyright (c) 2016 John Wiley & Sons, Ltd.
Sequential Elimination (S.E.) is a simple approach, based on standard algorithms, to the sorting of large dimensional arrays for the estimation of quantiles of unknown distributions. sorting on sub-arrays and eliminat...
详细信息
Sequential Elimination (S.E.) is a simple approach, based on standard algorithms, to the sorting of large dimensional arrays for the estimation of quantiles of unknown distributions. sorting on sub-arrays and eliminating elements outside properly constructed intervals, S.E. is faster than available alternatives and produces unbiased, consistent and efficient estimates. S.E. is used to tabulate critical values for ADF and conditional EG from 1010 simulations for the testing of unitroots and no-cointegration, respectively. The new critical values are applied to the testing of the presence of rational bubbles in the U.S. stock market.
The main idea of Optimized Selection Sort Algorithm (OSSA) is based on the already existing selection sort algorithm, with a difference that old selection sort;sorts one element either smallest or largest in a single ...
详细信息
ISBN:
(纸本)9781467376839
The main idea of Optimized Selection Sort Algorithm (OSSA) is based on the already existing selection sort algorithm, with a difference that old selection sort;sorts one element either smallest or largest in a single iteration while optimized selection sort, sorts both the elements at the same time i.e smallest and largest in a single iteration. In this study we have developed a variation of OSSA for two-dimensional array and called it Optimized Selection Sort algorithms for Two-Dimensional arrays OSSA2D. The hypothetical and experimental analysis revealed that the implementation of the proposed algorithm is easy. The comparison shows that the performance of OSSA2D is better than OSSA by four times and when compared with old Selection Sort algorithm the performance is improved by eight times (i.e if OSSA can sort an array in 100 seconds, OSSA2D can sort it in 24.55 Seconds, and similarly if Selection Sort takes 100 Seconds then OSSA2D take only 12.22 Seconds). This performance is remarkable when the array size is very large. The experiential results also demonstrate that the proposed algorithm has much lower computational complexity than the one dimensional sorting algorithm when the array size is very large.
暂无评论