Robust Multivariate Regression with Grossly Corrupted Observations and Its Application to Personality Prediction

Xiaowei Zhang, Li Cheng, Tingshao Zhu
; Asian Conference on Machine Learning, PMLR 45:112-126, 2016.

Abstract

We consider the problem of multivariate linear regression with a small fraction of the responses being missing and grossly corrupted, where the magnitudes and locations of such occurrences are not known in priori. This is addressed in our approach by explicitly taking into account the error source and its sparseness nature. Moreover, our approach allows each regression task to possess its distinct noise level. We also propose a new algorithm that is theoretically shown to always converge to the optimal solution of its induced non-smooth optimization problem. Experiments on controlled simulations suggest the competitiveness of our algorithm comparing to existing multivariate regression models. In particular, we apply our model to predict the \textitBig-Five personality from user behaviors at Social Network Sites (SNSs) and microblogs, an important yet difficult problem in psychology, where empirical results demonstrate its superior performance with respect to related learning methods.

Cite this Paper


BibTeX
@InProceedings{pmlr-v45-Zhang15a, title = {Robust Multivariate Regression with Grossly Corrupted Observations and Its Application to Personality Prediction}, author = {Xiaowei Zhang and Li Cheng and Tingshao Zhu}, booktitle = {Asian Conference on Machine Learning}, pages = {112--126}, year = {2016}, editor = {Geoffrey Holmes and Tie-Yan Liu}, volume = {45}, series = {Proceedings of Machine Learning Research}, address = {Hong Kong}, month = {20--22 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v45/Zhang15a.pdf}, url = {http://proceedings.mlr.press/v45/Zhang15a.html}, abstract = {We consider the problem of multivariate linear regression with a small fraction of the responses being missing and grossly corrupted, where the magnitudes and locations of such occurrences are not known in priori. This is addressed in our approach by explicitly taking into account the error source and its sparseness nature. Moreover, our approach allows each regression task to possess its distinct noise level. We also propose a new algorithm that is theoretically shown to always converge to the optimal solution of its induced non-smooth optimization problem. Experiments on controlled simulations suggest the competitiveness of our algorithm comparing to existing multivariate regression models. In particular, we apply our model to predict the \textitBig-Five personality from user behaviors at Social Network Sites (SNSs) and microblogs, an important yet difficult problem in psychology, where empirical results demonstrate its superior performance with respect to related learning methods. } }
Endnote
%0 Conference Paper %T Robust Multivariate Regression with Grossly Corrupted Observations and Its Application to Personality Prediction %A Xiaowei Zhang %A Li Cheng %A Tingshao Zhu %B Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2016 %E Geoffrey Holmes %E Tie-Yan Liu %F pmlr-v45-Zhang15a %I PMLR %J Proceedings of Machine Learning Research %P 112--126 %U http://proceedings.mlr.press %V 45 %W PMLR %X We consider the problem of multivariate linear regression with a small fraction of the responses being missing and grossly corrupted, where the magnitudes and locations of such occurrences are not known in priori. This is addressed in our approach by explicitly taking into account the error source and its sparseness nature. Moreover, our approach allows each regression task to possess its distinct noise level. We also propose a new algorithm that is theoretically shown to always converge to the optimal solution of its induced non-smooth optimization problem. Experiments on controlled simulations suggest the competitiveness of our algorithm comparing to existing multivariate regression models. In particular, we apply our model to predict the \textitBig-Five personality from user behaviors at Social Network Sites (SNSs) and microblogs, an important yet difficult problem in psychology, where empirical results demonstrate its superior performance with respect to related learning methods.
RIS
TY - CPAPER TI - Robust Multivariate Regression with Grossly Corrupted Observations and Its Application to Personality Prediction AU - Xiaowei Zhang AU - Li Cheng AU - Tingshao Zhu BT - Asian Conference on Machine Learning PY - 2016/02/25 DA - 2016/02/25 ED - Geoffrey Holmes ED - Tie-Yan Liu ID - pmlr-v45-Zhang15a PB - PMLR SP - 112 DP - PMLR EP - 126 L1 - http://proceedings.mlr.press/v45/Zhang15a.pdf UR - http://proceedings.mlr.press/v45/Zhang15a.html AB - We consider the problem of multivariate linear regression with a small fraction of the responses being missing and grossly corrupted, where the magnitudes and locations of such occurrences are not known in priori. This is addressed in our approach by explicitly taking into account the error source and its sparseness nature. Moreover, our approach allows each regression task to possess its distinct noise level. We also propose a new algorithm that is theoretically shown to always converge to the optimal solution of its induced non-smooth optimization problem. Experiments on controlled simulations suggest the competitiveness of our algorithm comparing to existing multivariate regression models. In particular, we apply our model to predict the \textitBig-Five personality from user behaviors at Social Network Sites (SNSs) and microblogs, an important yet difficult problem in psychology, where empirical results demonstrate its superior performance with respect to related learning methods. ER -
APA
Zhang, X., Cheng, L. & Zhu, T.. (2016). Robust Multivariate Regression with Grossly Corrupted Observations and Its Application to Personality Prediction. Asian Conference on Machine Learning, in PMLR 45:112-126

Related Material