Greedy Bilateral Sketch, Completion & Smoothing

Tianyi Zhou, Dacheng Tao
Proceedings of the Sixteenth International Conference on Artificial Intelligence and Statistics, PMLR 31:650-658, 2013.

Abstract

Recovering a large low-rank matrix from highly corrupted, incomplete or sparse outlier overwhelmed observations is the crux of various intriguing statistical problems. We explore the power of "greedy bilateral (GreB)" paradigm in reducing both time and sample complexities for solving these problems. GreB models a low-rank variable as a bilateral factorization, and updates the left and right factors in a mutually adaptive and greedy incremental manner. We detail how to model and solve low-rank approximation, matrix completion and robust PCA in GreB’s paradigm. On their MATLAB implementations, approximating a noisy 10000x10000 matrix of rank 500 with SVD accuracy takes 6s; MovieLens10M matrix of size 69878x10677 can be completed in 10s from 30% of 10^7 ratings with RMSE 0.86 on the rest 70%; the low-rank background and sparse moving outliers in a 120x160 video of 500 frames are accurately separated in 1s. This brings 30 to 100 times acceleration in solving these popular statistical problems.

Cite this Paper


BibTeX
@InProceedings{pmlr-v31-zhou13b, title = {Greedy Bilateral Sketch, Completion & Smoothing}, author = {Zhou, Tianyi and Tao, Dacheng}, booktitle = {Proceedings of the Sixteenth International Conference on Artificial Intelligence and Statistics}, pages = {650--658}, year = {2013}, editor = {Carvalho, Carlos M. and Ravikumar, Pradeep}, volume = {31}, series = {Proceedings of Machine Learning Research}, address = {Scottsdale, Arizona, USA}, month = {29 Apr--01 May}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v31/zhou13b.pdf}, url = {https://proceedings.mlr.press/v31/zhou13b.html}, abstract = {Recovering a large low-rank matrix from highly corrupted, incomplete or sparse outlier overwhelmed observations is the crux of various intriguing statistical problems. We explore the power of "greedy bilateral (GreB)" paradigm in reducing both time and sample complexities for solving these problems. GreB models a low-rank variable as a bilateral factorization, and updates the left and right factors in a mutually adaptive and greedy incremental manner. We detail how to model and solve low-rank approximation, matrix completion and robust PCA in GreB’s paradigm. On their MATLAB implementations, approximating a noisy 10000x10000 matrix of rank 500 with SVD accuracy takes 6s; MovieLens10M matrix of size 69878x10677 can be completed in 10s from 30% of 10^7 ratings with RMSE 0.86 on the rest 70%; the low-rank background and sparse moving outliers in a 120x160 video of 500 frames are accurately separated in 1s. This brings 30 to 100 times acceleration in solving these popular statistical problems.} }
Endnote
%0 Conference Paper %T Greedy Bilateral Sketch, Completion & Smoothing %A Tianyi Zhou %A Dacheng Tao %B Proceedings of the Sixteenth International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2013 %E Carlos M. Carvalho %E Pradeep Ravikumar %F pmlr-v31-zhou13b %I PMLR %P 650--658 %U https://proceedings.mlr.press/v31/zhou13b.html %V 31 %X Recovering a large low-rank matrix from highly corrupted, incomplete or sparse outlier overwhelmed observations is the crux of various intriguing statistical problems. We explore the power of "greedy bilateral (GreB)" paradigm in reducing both time and sample complexities for solving these problems. GreB models a low-rank variable as a bilateral factorization, and updates the left and right factors in a mutually adaptive and greedy incremental manner. We detail how to model and solve low-rank approximation, matrix completion and robust PCA in GreB’s paradigm. On their MATLAB implementations, approximating a noisy 10000x10000 matrix of rank 500 with SVD accuracy takes 6s; MovieLens10M matrix of size 69878x10677 can be completed in 10s from 30% of 10^7 ratings with RMSE 0.86 on the rest 70%; the low-rank background and sparse moving outliers in a 120x160 video of 500 frames are accurately separated in 1s. This brings 30 to 100 times acceleration in solving these popular statistical problems.
RIS
TY - CPAPER TI - Greedy Bilateral Sketch, Completion & Smoothing AU - Tianyi Zhou AU - Dacheng Tao BT - Proceedings of the Sixteenth International Conference on Artificial Intelligence and Statistics DA - 2013/04/29 ED - Carlos M. Carvalho ED - Pradeep Ravikumar ID - pmlr-v31-zhou13b PB - PMLR DP - Proceedings of Machine Learning Research VL - 31 SP - 650 EP - 658 L1 - http://proceedings.mlr.press/v31/zhou13b.pdf UR - https://proceedings.mlr.press/v31/zhou13b.html AB - Recovering a large low-rank matrix from highly corrupted, incomplete or sparse outlier overwhelmed observations is the crux of various intriguing statistical problems. We explore the power of "greedy bilateral (GreB)" paradigm in reducing both time and sample complexities for solving these problems. GreB models a low-rank variable as a bilateral factorization, and updates the left and right factors in a mutually adaptive and greedy incremental manner. We detail how to model and solve low-rank approximation, matrix completion and robust PCA in GreB’s paradigm. On their MATLAB implementations, approximating a noisy 10000x10000 matrix of rank 500 with SVD accuracy takes 6s; MovieLens10M matrix of size 69878x10677 can be completed in 10s from 30% of 10^7 ratings with RMSE 0.86 on the rest 70%; the low-rank background and sparse moving outliers in a 120x160 video of 500 frames are accurately separated in 1s. This brings 30 to 100 times acceleration in solving these popular statistical problems. ER -
APA
Zhou, T. & Tao, D.. (2013). Greedy Bilateral Sketch, Completion & Smoothing. Proceedings of the Sixteenth International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 31:650-658 Available from https://proceedings.mlr.press/v31/zhou13b.html.

Related Material