An Online Learning Algorithm for Bilinear Models

Yuanbin Wu, Shiliang Sun
Proceedings of the 32nd International Conference on Machine Learning, PMLR 37:890-898, 2015.

Abstract

We investigate the bilinear model, which is a matrix form linear model with the rank 1 constraint. A new online learning algorithm is proposed to train the model parameters. Our algorithm runs in the manner of online mirror descent, and gradients are computed by the power iteration. To analyze it, we give a new second order approximation of the squared spectral norm, which helps us to get a regret bound. Experiments on two sequential labelling tasks give positive results.

Cite this Paper


BibTeX
@InProceedings{pmlr-v37-wua15, title = {An Online Learning Algorithm for Bilinear Models}, author = {Wu, Yuanbin and Sun, Shiliang}, booktitle = {Proceedings of the 32nd International Conference on Machine Learning}, pages = {890--898}, year = {2015}, editor = {Bach, Francis and Blei, David}, volume = {37}, series = {Proceedings of Machine Learning Research}, address = {Lille, France}, month = {07--09 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v37/wua15.pdf}, url = {https://proceedings.mlr.press/v37/wua15.html}, abstract = {We investigate the bilinear model, which is a matrix form linear model with the rank 1 constraint. A new online learning algorithm is proposed to train the model parameters. Our algorithm runs in the manner of online mirror descent, and gradients are computed by the power iteration. To analyze it, we give a new second order approximation of the squared spectral norm, which helps us to get a regret bound. Experiments on two sequential labelling tasks give positive results.} }
Endnote
%0 Conference Paper %T An Online Learning Algorithm for Bilinear Models %A Yuanbin Wu %A Shiliang Sun %B Proceedings of the 32nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2015 %E Francis Bach %E David Blei %F pmlr-v37-wua15 %I PMLR %P 890--898 %U https://proceedings.mlr.press/v37/wua15.html %V 37 %X We investigate the bilinear model, which is a matrix form linear model with the rank 1 constraint. A new online learning algorithm is proposed to train the model parameters. Our algorithm runs in the manner of online mirror descent, and gradients are computed by the power iteration. To analyze it, we give a new second order approximation of the squared spectral norm, which helps us to get a regret bound. Experiments on two sequential labelling tasks give positive results.
RIS
TY - CPAPER TI - An Online Learning Algorithm for Bilinear Models AU - Yuanbin Wu AU - Shiliang Sun BT - Proceedings of the 32nd International Conference on Machine Learning DA - 2015/06/01 ED - Francis Bach ED - David Blei ID - pmlr-v37-wua15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 37 SP - 890 EP - 898 L1 - http://proceedings.mlr.press/v37/wua15.pdf UR - https://proceedings.mlr.press/v37/wua15.html AB - We investigate the bilinear model, which is a matrix form linear model with the rank 1 constraint. A new online learning algorithm is proposed to train the model parameters. Our algorithm runs in the manner of online mirror descent, and gradients are computed by the power iteration. To analyze it, we give a new second order approximation of the squared spectral norm, which helps us to get a regret bound. Experiments on two sequential labelling tasks give positive results. ER -
APA
Wu, Y. & Sun, S.. (2015). An Online Learning Algorithm for Bilinear Models. Proceedings of the 32nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 37:890-898 Available from https://proceedings.mlr.press/v37/wua15.html.

Related Material