Accelerated Online Low Rank Tensor Learning for Multivariate Spatiotemporal Streams

Rose Yu, Dehua Cheng, Yan Liu
Proceedings of the 32nd International Conference on Machine Learning, PMLR 37:238-247, 2015.

Abstract

Low-rank tensor learning has many applications in machine learning. A series of batch learning algorithms have achieved great successes. However, in many emerging applications, such as climate data analysis, we are confronted with large-scale tensor streams, which poses significant challenges to existing solution in terms of computational costs and limited response time. In this paper, we propose an online accelerated low-rank tensor learning algorithm (ALTO) to solve the problem. At each iteration, we project the current tensor to the subspace of low-rank tensors in order to perform efficient tensor decomposition, then recover the decomposition of the new tensor. By randomly glancing at additional subspaces, we successfully avoid local optima at negligible extra computational cost. We evaluate our method on two tasks in streaming multivariate spatio-temporal analysis: online forecasting and multi-model ensemble, which shows that our method achieves comparable predictive accuracy with significant boost in run time.

Cite this Paper


BibTeX
@InProceedings{pmlr-v37-yua15, title = {Accelerated Online Low Rank Tensor Learning for Multivariate Spatiotemporal Streams}, author = {Yu, Rose and Cheng, Dehua and Liu, Yan}, booktitle = {Proceedings of the 32nd International Conference on Machine Learning}, pages = {238--247}, 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/yua15.pdf}, url = {https://proceedings.mlr.press/v37/yua15.html}, abstract = {Low-rank tensor learning has many applications in machine learning. A series of batch learning algorithms have achieved great successes. However, in many emerging applications, such as climate data analysis, we are confronted with large-scale tensor streams, which poses significant challenges to existing solution in terms of computational costs and limited response time. In this paper, we propose an online accelerated low-rank tensor learning algorithm (ALTO) to solve the problem. At each iteration, we project the current tensor to the subspace of low-rank tensors in order to perform efficient tensor decomposition, then recover the decomposition of the new tensor. By randomly glancing at additional subspaces, we successfully avoid local optima at negligible extra computational cost. We evaluate our method on two tasks in streaming multivariate spatio-temporal analysis: online forecasting and multi-model ensemble, which shows that our method achieves comparable predictive accuracy with significant boost in run time.} }
Endnote
%0 Conference Paper %T Accelerated Online Low Rank Tensor Learning for Multivariate Spatiotemporal Streams %A Rose Yu %A Dehua Cheng %A Yan Liu %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-yua15 %I PMLR %P 238--247 %U https://proceedings.mlr.press/v37/yua15.html %V 37 %X Low-rank tensor learning has many applications in machine learning. A series of batch learning algorithms have achieved great successes. However, in many emerging applications, such as climate data analysis, we are confronted with large-scale tensor streams, which poses significant challenges to existing solution in terms of computational costs and limited response time. In this paper, we propose an online accelerated low-rank tensor learning algorithm (ALTO) to solve the problem. At each iteration, we project the current tensor to the subspace of low-rank tensors in order to perform efficient tensor decomposition, then recover the decomposition of the new tensor. By randomly glancing at additional subspaces, we successfully avoid local optima at negligible extra computational cost. We evaluate our method on two tasks in streaming multivariate spatio-temporal analysis: online forecasting and multi-model ensemble, which shows that our method achieves comparable predictive accuracy with significant boost in run time.
RIS
TY - CPAPER TI - Accelerated Online Low Rank Tensor Learning for Multivariate Spatiotemporal Streams AU - Rose Yu AU - Dehua Cheng AU - Yan Liu BT - Proceedings of the 32nd International Conference on Machine Learning DA - 2015/06/01 ED - Francis Bach ED - David Blei ID - pmlr-v37-yua15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 37 SP - 238 EP - 247 L1 - http://proceedings.mlr.press/v37/yua15.pdf UR - https://proceedings.mlr.press/v37/yua15.html AB - Low-rank tensor learning has many applications in machine learning. A series of batch learning algorithms have achieved great successes. However, in many emerging applications, such as climate data analysis, we are confronted with large-scale tensor streams, which poses significant challenges to existing solution in terms of computational costs and limited response time. In this paper, we propose an online accelerated low-rank tensor learning algorithm (ALTO) to solve the problem. At each iteration, we project the current tensor to the subspace of low-rank tensors in order to perform efficient tensor decomposition, then recover the decomposition of the new tensor. By randomly glancing at additional subspaces, we successfully avoid local optima at negligible extra computational cost. We evaluate our method on two tasks in streaming multivariate spatio-temporal analysis: online forecasting and multi-model ensemble, which shows that our method achieves comparable predictive accuracy with significant boost in run time. ER -
APA
Yu, R., Cheng, D. & Liu, Y.. (2015). Accelerated Online Low Rank Tensor Learning for Multivariate Spatiotemporal Streams. Proceedings of the 32nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 37:238-247 Available from https://proceedings.mlr.press/v37/yua15.html.

Related Material