Local Deep Kernel Learning for Efficient Non-linear SVM Prediction

Cijo Jose, Prasoon Goyal, Parv Aggrwal, Manik Varma
Proceedings of the 30th International Conference on Machine Learning, PMLR 28(3):486-494, 2013.

Abstract

Our objective is to speed up non-linear SVM prediction while maintaining classification accuracy above an acceptable limit. We generalize Localized Multiple Kernel Learning so as to learn a primal feature space embedding which is high dimensional, sparse and computationally deep. Primal based classification decouples prediction costs from the number of support vectors and our tree-structured features efficiently encode non-linearities while speeding up prediction exponentially over the state-of-the-art. We develop routines for optimizing over the space of tree-structured features and efficiently scale to problems with over half a million training points. Experiments on benchmark data sets reveal that our formulation can reduce prediction costs by more than three orders of magnitude in some cases with a moderate sacrifice in classification accuracy as compared to RBF-SVMs. Furthermore, our formulation leads to much better classification accuracies over leading methods.

Cite this Paper

BibTeX
@InProceedings{pmlr-v28-jose13, title = {Local Deep Kernel Learning for Efficient Non-linear SVM Prediction}, author = {Jose, Cijo and Goyal, Prasoon and Aggrwal, Parv and Varma, Manik}, booktitle = {Proceedings of the 30th International Conference on Machine Learning}, pages = {486--494}, year = {2013}, editor = {Dasgupta, Sanjoy and McAllester, David}, volume = {28}, number = {3}, series = {Proceedings of Machine Learning Research}, address = {Atlanta, Georgia, USA}, month = {17--19 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v28/jose13.pdf}, url = {https://proceedings.mlr.press/v28/jose13.html}, abstract = {Our objective is to speed up non-linear SVM prediction while maintaining classification accuracy above an acceptable limit. We generalize Localized Multiple Kernel Learning so as to learn a primal feature space embedding which is high dimensional, sparse and computationally deep. Primal based classification decouples prediction costs from the number of support vectors and our tree-structured features efficiently encode non-linearities while speeding up prediction exponentially over the state-of-the-art. We develop routines for optimizing over the space of tree-structured features and efficiently scale to problems with over half a million training points. Experiments on benchmark data sets reveal that our formulation can reduce prediction costs by more than three orders of magnitude in some cases with a moderate sacrifice in classification accuracy as compared to RBF-SVMs. Furthermore, our formulation leads to much better classification accuracies over leading methods.} }
Endnote
%0 Conference Paper %T Local Deep Kernel Learning for Efficient Non-linear SVM Prediction %A Cijo Jose %A Prasoon Goyal %A Parv Aggrwal %A Manik Varma %B Proceedings of the 30th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2013 %E Sanjoy Dasgupta %E David McAllester %F pmlr-v28-jose13 %I PMLR %P 486--494 %U https://proceedings.mlr.press/v28/jose13.html %V 28 %N 3 %X Our objective is to speed up non-linear SVM prediction while maintaining classification accuracy above an acceptable limit. We generalize Localized Multiple Kernel Learning so as to learn a primal feature space embedding which is high dimensional, sparse and computationally deep. Primal based classification decouples prediction costs from the number of support vectors and our tree-structured features efficiently encode non-linearities while speeding up prediction exponentially over the state-of-the-art. We develop routines for optimizing over the space of tree-structured features and efficiently scale to problems with over half a million training points. Experiments on benchmark data sets reveal that our formulation can reduce prediction costs by more than three orders of magnitude in some cases with a moderate sacrifice in classification accuracy as compared to RBF-SVMs. Furthermore, our formulation leads to much better classification accuracies over leading methods.
RIS
TY - CPAPER TI - Local Deep Kernel Learning for Efficient Non-linear SVM Prediction AU - Cijo Jose AU - Prasoon Goyal AU - Parv Aggrwal AU - Manik Varma BT - Proceedings of the 30th International Conference on Machine Learning DA - 2013/05/26 ED - Sanjoy Dasgupta ED - David McAllester ID - pmlr-v28-jose13 PB - PMLR DP - Proceedings of Machine Learning Research VL - 28 IS - 3 SP - 486 EP - 494 L1 - http://proceedings.mlr.press/v28/jose13.pdf UR - https://proceedings.mlr.press/v28/jose13.html AB - Our objective is to speed up non-linear SVM prediction while maintaining classification accuracy above an acceptable limit. We generalize Localized Multiple Kernel Learning so as to learn a primal feature space embedding which is high dimensional, sparse and computationally deep. Primal based classification decouples prediction costs from the number of support vectors and our tree-structured features efficiently encode non-linearities while speeding up prediction exponentially over the state-of-the-art. We develop routines for optimizing over the space of tree-structured features and efficiently scale to problems with over half a million training points. Experiments on benchmark data sets reveal that our formulation can reduce prediction costs by more than three orders of magnitude in some cases with a moderate sacrifice in classification accuracy as compared to RBF-SVMs. Furthermore, our formulation leads to much better classification accuracies over leading methods. ER -
APA
Jose, C., Goyal, P., Aggrwal, P. & Varma, M.. (2013). Local Deep Kernel Learning for Efficient Non-linear SVM Prediction. Proceedings of the 30th International Conference on Machine Learning, in Proceedings of Machine Learning Research 28(3):486-494 Available from https://proceedings.mlr.press/v28/jose13.html.

Related Material