Back to the Future: Radial Basis Function Networks Revisited

Qichao Que, Mikhail Belkin
Proceedings of the 19th International Conference on Artificial Intelligence and Statistics, PMLR 51:1375-1383, 2016.

Abstract

Radial Basis Function (RBF) networks are a classical family of algorithms for supervised learning. The most popular approach for training RBF networks has relied on kernel methods using regularization based on a norm in a Reproducing Kernel Hilbert Space (RKHS), which is a principled and empirically successful framework. In this paper we aim to revisit some of the older approaches to training the RBF networks from a more modern perspective. Specifically, we analyze two common regularization procedures, one based on the square norm of the coefficients in the network and another on using centers obtained by k-means clustering. We show that both of these RBF methods can be recast as certain data-dependent kernels. We provide a theoretical analysis of these methods as well as a number of experimental results, pointing out very competitive experimental performance as well as certain advantages over the standard kernel methods in terms of both flexibility (incorporating of unlabeled data) and computational complexity. Finally, our results shed light on some impressive recent successes of using soft k-means features for image recognition and other tasks.

Cite this Paper


BibTeX
@InProceedings{pmlr-v51-que16, title = {Back to the Future: Radial Basis Function Networks Revisited}, author = {Que, Qichao and Belkin, Mikhail}, booktitle = {Proceedings of the 19th International Conference on Artificial Intelligence and Statistics}, pages = {1375--1383}, year = {2016}, editor = {Gretton, Arthur and Robert, Christian C.}, volume = {51}, series = {Proceedings of Machine Learning Research}, address = {Cadiz, Spain}, month = {09--11 May}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v51/que16.pdf}, url = {http://proceedings.mlr.press/v51/que16.html}, abstract = {Radial Basis Function (RBF) networks are a classical family of algorithms for supervised learning. The most popular approach for training RBF networks has relied on kernel methods using regularization based on a norm in a Reproducing Kernel Hilbert Space (RKHS), which is a principled and empirically successful framework. In this paper we aim to revisit some of the older approaches to training the RBF networks from a more modern perspective. Specifically, we analyze two common regularization procedures, one based on the square norm of the coefficients in the network and another on using centers obtained by k-means clustering. We show that both of these RBF methods can be recast as certain data-dependent kernels. We provide a theoretical analysis of these methods as well as a number of experimental results, pointing out very competitive experimental performance as well as certain advantages over the standard kernel methods in terms of both flexibility (incorporating of unlabeled data) and computational complexity. Finally, our results shed light on some impressive recent successes of using soft k-means features for image recognition and other tasks.} }
Endnote
%0 Conference Paper %T Back to the Future: Radial Basis Function Networks Revisited %A Qichao Que %A Mikhail Belkin %B Proceedings of the 19th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2016 %E Arthur Gretton %E Christian C. Robert %F pmlr-v51-que16 %I PMLR %P 1375--1383 %U http://proceedings.mlr.press/v51/que16.html %V 51 %X Radial Basis Function (RBF) networks are a classical family of algorithms for supervised learning. The most popular approach for training RBF networks has relied on kernel methods using regularization based on a norm in a Reproducing Kernel Hilbert Space (RKHS), which is a principled and empirically successful framework. In this paper we aim to revisit some of the older approaches to training the RBF networks from a more modern perspective. Specifically, we analyze two common regularization procedures, one based on the square norm of the coefficients in the network and another on using centers obtained by k-means clustering. We show that both of these RBF methods can be recast as certain data-dependent kernels. We provide a theoretical analysis of these methods as well as a number of experimental results, pointing out very competitive experimental performance as well as certain advantages over the standard kernel methods in terms of both flexibility (incorporating of unlabeled data) and computational complexity. Finally, our results shed light on some impressive recent successes of using soft k-means features for image recognition and other tasks.
RIS
TY - CPAPER TI - Back to the Future: Radial Basis Function Networks Revisited AU - Qichao Que AU - Mikhail Belkin BT - Proceedings of the 19th International Conference on Artificial Intelligence and Statistics DA - 2016/05/02 ED - Arthur Gretton ED - Christian C. Robert ID - pmlr-v51-que16 PB - PMLR DP - Proceedings of Machine Learning Research VL - 51 SP - 1375 EP - 1383 L1 - http://proceedings.mlr.press/v51/que16.pdf UR - http://proceedings.mlr.press/v51/que16.html AB - Radial Basis Function (RBF) networks are a classical family of algorithms for supervised learning. The most popular approach for training RBF networks has relied on kernel methods using regularization based on a norm in a Reproducing Kernel Hilbert Space (RKHS), which is a principled and empirically successful framework. In this paper we aim to revisit some of the older approaches to training the RBF networks from a more modern perspective. Specifically, we analyze two common regularization procedures, one based on the square norm of the coefficients in the network and another on using centers obtained by k-means clustering. We show that both of these RBF methods can be recast as certain data-dependent kernels. We provide a theoretical analysis of these methods as well as a number of experimental results, pointing out very competitive experimental performance as well as certain advantages over the standard kernel methods in terms of both flexibility (incorporating of unlabeled data) and computational complexity. Finally, our results shed light on some impressive recent successes of using soft k-means features for image recognition and other tasks. ER -
APA
Que, Q. & Belkin, M.. (2016). Back to the Future: Radial Basis Function Networks Revisited. Proceedings of the 19th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 51:1375-1383 Available from http://proceedings.mlr.press/v51/que16.html.

Related Material