Efficient Learning of Mahalanobis Metrics for Ranking
Daryl Lim, Gert R. G. Lanckriet · 2014
We develop an efficient algorithm to learn a Ma-halanobis distance metric by directly optimizing a ranking loss. Our approach focuses on optimiz-ing the top of the induced rankings, which is de-sirable in tasks such as visualization and nearest-neighbor retrieval. We further develop and justify a simple technique to reduce training time sig-nificantly with minimal impact on performance. Our proposed method significantly outperforms alternative methods on several real-world tasks, and can scale to large and high-dimensional data. 1.