论文信息 - Parallelization of Error Weighted Hashing for Approximate k-Nearest Neighbour Search on GPU-CPU Hybrid

Parallelization of Error Weighted Hashing for Approximate k-Nearest Neighbour Search on GPU-CPU Hybrid

Error Weighted Hashing (EWH) is a fast algorithm for Approximate k-Nearest neighbour search in Hamming space. It is more efficient than traditional LocalitySensitive Hashing algorithm (LSH) since it generates shorterlist of strings for finding the exact distance from the query. Wehave parallelized the EWH algorithm using Cuda and OpenMP.Speedup of 44 times on a 16 core GPU and 16 core CPUmachine was achieved in case for hashing and 24 times for retrieval.

Siddharth Bhatia | Mohan Pavan Kumar Badarla

[1] Ashish Goel,et al. Efficient distributed locality sensitive hashing , 2012, CIKM.

[2] Rabab Kreidieh Ward,et al. A Fast Approximate Nearest Neighbor Search Algorithm in the Hamming Space , 2012, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[3] Dinesh Manocha,et al. Fast GPU-based locality sensitive hashing for k-nearest neighbor computation , 2011, GIS.

[4] Andrew W. Moore,et al. An Investigation of Practical Approximate Nearest Neighbor Algorithms , 2004, NIPS.

[5] Jeremy Buhler,et al. Efficient large-scale sequence comparison by locality-sensitive hashing , 2001, Bioinform..

[6] Pawan Gupta,et al. Distributed Locality Sensitivity Hashing , 2010, 2010 7th IEEE Consumer Communications and Networking Conference.