Truncated Quantization for Approximate Nearest Neighbor Search
碩士 === 國立清華大學 === 資訊工程學系 === 103 === We introduce a new observation about vector quantization and dimensionality reduction. This observation can help to improve the quality of approximate nearest neighbor search. Based on the observation we develop an efficient algorithm leveraging the quantization...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | en_US |
| Published: |
2015
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/77385632500347288796 |
| Summary: | 碩士 === 國立清華大學 === 資訊工程學系 === 103 === We introduce a new observation about vector quantization and dimensionality reduction. This observation can help to improve the quality of approximate nearest neighbor search. Based on the observation we develop an efficient algorithm leveraging the quantization error and the balanced variances of subspaces criteria for codebook learning. Experimental results show that our approach is able to achieve better performance on searching large datasets. We also present an application that takes advantage of fast approximate nearest neighbor search with our approach.
|
|---|