Fastfood - Computing Hilbert Space Expansions in loglinear time

Quoc Le, Tamas Sarlos, Alexander Smola
Proceedings of the 30th International Conference on Machine Learning, PMLR 28(3):244-252, 2013.

Abstract

Fast nonlinear function classes are crucial for nonparametric estimation, such as in kernel methods. This paper proposes an improvement to random kitchen sinks that offers significantly faster computation in log-linear time without sacrificing accuracy. Furthermore, we show how one may adjust the regularization properties of the kernel simply by changing the spectral distribution of the projection matrix. We provide experimental results which show that even for for moderately small problems we already achieve two orders of magnitude faster computation and three orders of magnitude lower memory footprint.

Cite this Paper

BibTeX
@InProceedings{pmlr-v28-le13, title = {Fastfood - Computing Hilbert Space Expansions in loglinear time}, author = {Le, Quoc and Sarlos, Tamas and Smola, Alexander}, booktitle = {Proceedings of the 30th International Conference on Machine Learning}, pages = {244--252}, 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/le13.pdf}, url = {https://proceedings.mlr.press/v28/le13.html}, abstract = {Fast nonlinear function classes are crucial for nonparametric estimation, such as in kernel methods. This paper proposes an improvement to random kitchen sinks that offers significantly faster computation in log-linear time without sacrificing accuracy. Furthermore, we show how one may adjust the regularization properties of the kernel simply by changing the spectral distribution of the projection matrix. We provide experimental results which show that even for for moderately small problems we already achieve two orders of magnitude faster computation and three orders of magnitude lower memory footprint. } }
Endnote
%0 Conference Paper %T Fastfood - Computing Hilbert Space Expansions in loglinear time %A Quoc Le %A Tamas Sarlos %A Alexander Smola %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-le13 %I PMLR %P 244--252 %U https://proceedings.mlr.press/v28/le13.html %V 28 %N 3 %X Fast nonlinear function classes are crucial for nonparametric estimation, such as in kernel methods. This paper proposes an improvement to random kitchen sinks that offers significantly faster computation in log-linear time without sacrificing accuracy. Furthermore, we show how one may adjust the regularization properties of the kernel simply by changing the spectral distribution of the projection matrix. We provide experimental results which show that even for for moderately small problems we already achieve two orders of magnitude faster computation and three orders of magnitude lower memory footprint.
RIS
TY - CPAPER TI - Fastfood - Computing Hilbert Space Expansions in loglinear time AU - Quoc Le AU - Tamas Sarlos AU - Alexander Smola BT - Proceedings of the 30th International Conference on Machine Learning DA - 2013/05/26 ED - Sanjoy Dasgupta ED - David McAllester ID - pmlr-v28-le13 PB - PMLR DP - Proceedings of Machine Learning Research VL - 28 IS - 3 SP - 244 EP - 252 L1 - http://proceedings.mlr.press/v28/le13.pdf UR - https://proceedings.mlr.press/v28/le13.html AB - Fast nonlinear function classes are crucial for nonparametric estimation, such as in kernel methods. This paper proposes an improvement to random kitchen sinks that offers significantly faster computation in log-linear time without sacrificing accuracy. Furthermore, we show how one may adjust the regularization properties of the kernel simply by changing the spectral distribution of the projection matrix. We provide experimental results which show that even for for moderately small problems we already achieve two orders of magnitude faster computation and three orders of magnitude lower memory footprint. ER -
APA
Le, Q., Sarlos, T. & Smola, A.. (2013). Fastfood - Computing Hilbert Space Expansions in loglinear time. Proceedings of the 30th International Conference on Machine Learning, in Proceedings of Machine Learning Research 28(3):244-252 Available from https://proceedings.mlr.press/v28/le13.html.

Related Material