Efficient and scalable approaches to Mahalanobis distance metric learning
The development of an appropriate data-dependent distance metric is a compelling goal for many visual recognition tasks. This thesis proposes three efficient and scalable distance learning algorithms by employing the principle of margin maximization to secure better generalization performances. The proposed algorithms formulate metric learning as a convex optimization problem with a positive semidefinite (psd) matrix variable. A standard interior-point semidefinite programming (SDP) solver has...[Show more]
|Collections||Open Access Theses|
|Access Rights:||Open Access|
|b30880543-Kim_J.pdf||289.09 MB||Adobe PDF|
Items in Open Research are protected by copyright, with all rights reserved, unless otherwise indicated.