Generic Methods for Optimization-Based Modeling

Justin Domke
Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics, PMLR 22:318-326, 2012.

Abstract

"Energy” models for continuous domains can be applied to many problems, but often suffer from high computational expense in training, due to the need to repeatedly minimize the energy function to high accuracy. This paper considers a modified setting, where the model is trained in terms of results after optimization is truncated to a fixed number of iterations. We derive “backpropagating” versions of gradient descent, heavy-ball and LBFGS. These are simple to use, as they require as input only routines to compute the gradient of the energy with respect to the domain and parameters. Experimental results on denoising and image labeling problems show that learning with truncated optimization greatly reduces computational expense compared to “full” fitting.

Cite this Paper


BibTeX
@InProceedings{pmlr-v22-domke12, title = {Generic Methods for Optimization-Based Modeling}, author = {Domke, Justin}, booktitle = {Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics}, pages = {318--326}, year = {2012}, editor = {Lawrence, Neil D. and Girolami, Mark}, volume = {22}, series = {Proceedings of Machine Learning Research}, address = {La Palma, Canary Islands}, month = {21--23 Apr}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v22/domke12/domke12.pdf}, url = {https://proceedings.mlr.press/v22/domke12.html}, abstract = {"Energy” models for continuous domains can be applied to many problems, but often suffer from high computational expense in training, due to the need to repeatedly minimize the energy function to high accuracy. This paper considers a modified setting, where the model is trained in terms of results after optimization is truncated to a fixed number of iterations. We derive “backpropagating” versions of gradient descent, heavy-ball and LBFGS. These are simple to use, as they require as input only routines to compute the gradient of the energy with respect to the domain and parameters. Experimental results on denoising and image labeling problems show that learning with truncated optimization greatly reduces computational expense compared to “full” fitting.} }
Endnote
%0 Conference Paper %T Generic Methods for Optimization-Based Modeling %A Justin Domke %B Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2012 %E Neil D. Lawrence %E Mark Girolami %F pmlr-v22-domke12 %I PMLR %P 318--326 %U https://proceedings.mlr.press/v22/domke12.html %V 22 %X "Energy” models for continuous domains can be applied to many problems, but often suffer from high computational expense in training, due to the need to repeatedly minimize the energy function to high accuracy. This paper considers a modified setting, where the model is trained in terms of results after optimization is truncated to a fixed number of iterations. We derive “backpropagating” versions of gradient descent, heavy-ball and LBFGS. These are simple to use, as they require as input only routines to compute the gradient of the energy with respect to the domain and parameters. Experimental results on denoising and image labeling problems show that learning with truncated optimization greatly reduces computational expense compared to “full” fitting.
RIS
TY - CPAPER TI - Generic Methods for Optimization-Based Modeling AU - Justin Domke BT - Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics DA - 2012/03/21 ED - Neil D. Lawrence ED - Mark Girolami ID - pmlr-v22-domke12 PB - PMLR DP - Proceedings of Machine Learning Research VL - 22 SP - 318 EP - 326 L1 - http://proceedings.mlr.press/v22/domke12/domke12.pdf UR - https://proceedings.mlr.press/v22/domke12.html AB - "Energy” models for continuous domains can be applied to many problems, but often suffer from high computational expense in training, due to the need to repeatedly minimize the energy function to high accuracy. This paper considers a modified setting, where the model is trained in terms of results after optimization is truncated to a fixed number of iterations. We derive “backpropagating” versions of gradient descent, heavy-ball and LBFGS. These are simple to use, as they require as input only routines to compute the gradient of the energy with respect to the domain and parameters. Experimental results on denoising and image labeling problems show that learning with truncated optimization greatly reduces computational expense compared to “full” fitting. ER -
APA
Domke, J.. (2012). Generic Methods for Optimization-Based Modeling. Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 22:318-326 Available from https://proceedings.mlr.press/v22/domke12.html.

Related Material