GPU Acceleration of Similarity Search for Uncertain Time Series
Jun Hwang, Yusuke Kozawa, Toshiyuki Amagasa, Hiroyuki Kitagawa · 2014
Time series data often contain uncertainty due to various reasons, and the similarity search over uncertain time series data has been applied in many applications. For this reason, many methods have been proposed, and DUST is one of the latest methods that can deal with arbitrary probability distributions. However, it is known that its computational cost is high in particular when the dataset is large. To cope with this problem, in this paper, we attempt to improve the performance of DUST using GPU. More precisely, we speed up the computation by parallelizing the probability computation. The experimental evaluation reveals that the proposed scheme is much faster than the CPU-based implementation.