Learning in reproducing kernel Kreın spaces

We formulate a novel regularized risk minimization problem for learning in reproducing kernel Kreın spaces and show that the strong representer theorem applies to it. As a result of the latter, the learning problem can be expressed as the minimization of a quadratic form over a hypersphere of consta...

Full description

Bibliographic Details
Main Authors: Oglic, Dino, Gärtner, Thomas
Format: Conference or Workshop Item
Published: 2018
Online Access:https://eprints.nottingham.ac.uk/52412/
_version_ 1848798720400818176
author Oglic, Dino
Gärtner, Thomas
author_facet Oglic, Dino
Gärtner, Thomas
author_sort Oglic, Dino
building Nottingham Research Data Repository
collection Online Access
description We formulate a novel regularized risk minimization problem for learning in reproducing kernel Kreın spaces and show that the strong representer theorem applies to it. As a result of the latter, the learning problem can be expressed as the minimization of a quadratic form over a hypersphere of constant radius. We present an algorithm that can find a globally optimal solution to this nonconvex optimization problem in time cubic in the number of instances. Moreover, we derive the gradient of the solution with respect to its hyperparameters and, in this way, provide means for efficient hyperparameter tuning. The approach comes with a generalization bound expressed in terms of the Rademacher complexity of the corresponding hypothesis space. The major advantage over standard kernel methods is the ability to learn with various domain specific similarity measures for which positive definiteness does not hold or is difficult to establish. The approach is evaluated empirically using indefinite kernels defined on structured as well as vectorial data. The empirical results demonstrate a superior performance of our approach over the state-of-the-art baselines.
first_indexed 2025-11-14T20:24:15Z
format Conference or Workshop Item
id nottingham-52412
institution University of Nottingham Malaysia Campus
institution_category Local University
last_indexed 2025-11-14T20:24:15Z
publishDate 2018
recordtype eprints
repository_type Digital Repository
spelling nottingham-524122020-05-04T19:45:54Z https://eprints.nottingham.ac.uk/52412/ Learning in reproducing kernel Kreın spaces Oglic, Dino Gärtner, Thomas We formulate a novel regularized risk minimization problem for learning in reproducing kernel Kreın spaces and show that the strong representer theorem applies to it. As a result of the latter, the learning problem can be expressed as the minimization of a quadratic form over a hypersphere of constant radius. We present an algorithm that can find a globally optimal solution to this nonconvex optimization problem in time cubic in the number of instances. Moreover, we derive the gradient of the solution with respect to its hyperparameters and, in this way, provide means for efficient hyperparameter tuning. The approach comes with a generalization bound expressed in terms of the Rademacher complexity of the corresponding hypothesis space. The major advantage over standard kernel methods is the ability to learn with various domain specific similarity measures for which positive definiteness does not hold or is difficult to establish. The approach is evaluated empirically using indefinite kernels defined on structured as well as vectorial data. The empirical results demonstrate a superior performance of our approach over the state-of-the-art baselines. 2018-07-10 Conference or Workshop Item PeerReviewed Oglic, Dino and Gärtner, Thomas (2018) Learning in reproducing kernel Kreın spaces. In: 35th International Conference on Machine Learning, 10-15 July, 2018, Stockholm, Sweden.
spellingShingle Oglic, Dino
Gärtner, Thomas
Learning in reproducing kernel Kreın spaces
title Learning in reproducing kernel Kreın spaces
title_full Learning in reproducing kernel Kreın spaces
title_fullStr Learning in reproducing kernel Kreın spaces
title_full_unstemmed Learning in reproducing kernel Kreın spaces
title_short Learning in reproducing kernel Kreın spaces
title_sort learning in reproducing kernel kreın spaces
url https://eprints.nottingham.ac.uk/52412/