One-Pass Multi-View Learning

Yue Zhu, Wei Gao, Zhi-Hua Zhou
Asian Conference on Machine Learning, PMLR 45:407-422, 2016.

Abstract

Multi-view learning has been an important learning paradigm where data come from multiple channels or appear in multiple modalities. Many approaches have been developed in this field, and have achieved better performance than single-view ones. Those approaches, however, always work on small-size datasets with low dimensionality, owing to their high computational cost. In recent years, it has been witnessed that many applications involve large-scale multi-view data, e.g., hundreds of hours of video (including visual, audio and text views) is uploaded to YouTube every minute, bringing a big challenge to previous multi-view algorithms. This work concentrates on the large-scale multi-view learning for classification and proposes the One-Pass Multi-View (OPMV) framework which goes through the training data only once without storing the entire training examples. This approach jointly optimizes the composite objective functions with consistency linear constraints for different views. We verify, both theoretically and empirically, the effectiveness of the proposed algorithm.

Cite this Paper

BibTeX
@InProceedings{pmlr-v45-Zhu15, title = {One-Pass Multi-View Learning}, author = {Zhu, Yue and Gao, Wei and Zhou, Zhi-Hua}, booktitle = {Asian Conference on Machine Learning}, pages = {407--422}, year = {2016}, editor = {Holmes, Geoffrey and Liu, Tie-Yan}, volume = {45}, series = {Proceedings of Machine Learning Research}, address = {Hong Kong}, month = {20--22 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v45/Zhu15.pdf}, url = {https://proceedings.mlr.press/v45/Zhu15.html}, abstract = {Multi-view learning has been an important learning paradigm where data come from multiple channels or appear in multiple modalities. Many approaches have been developed in this field, and have achieved better performance than single-view ones. Those approaches, however, always work on small-size datasets with low dimensionality, owing to their high computational cost. In recent years, it has been witnessed that many applications involve large-scale multi-view data, e.g., hundreds of hours of video (including visual, audio and text views) is uploaded to YouTube every minute, bringing a big challenge to previous multi-view algorithms. This work concentrates on the large-scale multi-view learning for classification and proposes the One-Pass Multi-View (OPMV) framework which goes through the training data only once without storing the entire training examples. This approach jointly optimizes the composite objective functions with consistency linear constraints for different views. We verify, both theoretically and empirically, the effectiveness of the proposed algorithm. } }
Endnote
%0 Conference Paper %T One-Pass Multi-View Learning %A Yue Zhu %A Wei Gao %A Zhi-Hua Zhou %B Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2016 %E Geoffrey Holmes %E Tie-Yan Liu %F pmlr-v45-Zhu15 %I PMLR %P 407--422 %U https://proceedings.mlr.press/v45/Zhu15.html %V 45 %X Multi-view learning has been an important learning paradigm where data come from multiple channels or appear in multiple modalities. Many approaches have been developed in this field, and have achieved better performance than single-view ones. Those approaches, however, always work on small-size datasets with low dimensionality, owing to their high computational cost. In recent years, it has been witnessed that many applications involve large-scale multi-view data, e.g., hundreds of hours of video (including visual, audio and text views) is uploaded to YouTube every minute, bringing a big challenge to previous multi-view algorithms. This work concentrates on the large-scale multi-view learning for classification and proposes the One-Pass Multi-View (OPMV) framework which goes through the training data only once without storing the entire training examples. This approach jointly optimizes the composite objective functions with consistency linear constraints for different views. We verify, both theoretically and empirically, the effectiveness of the proposed algorithm.
RIS
TY - CPAPER TI - One-Pass Multi-View Learning AU - Yue Zhu AU - Wei Gao AU - Zhi-Hua Zhou BT - Asian Conference on Machine Learning DA - 2016/02/25 ED - Geoffrey Holmes ED - Tie-Yan Liu ID - pmlr-v45-Zhu15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 45 SP - 407 EP - 422 L1 - http://proceedings.mlr.press/v45/Zhu15.pdf UR - https://proceedings.mlr.press/v45/Zhu15.html AB - Multi-view learning has been an important learning paradigm where data come from multiple channels or appear in multiple modalities. Many approaches have been developed in this field, and have achieved better performance than single-view ones. Those approaches, however, always work on small-size datasets with low dimensionality, owing to their high computational cost. In recent years, it has been witnessed that many applications involve large-scale multi-view data, e.g., hundreds of hours of video (including visual, audio and text views) is uploaded to YouTube every minute, bringing a big challenge to previous multi-view algorithms. This work concentrates on the large-scale multi-view learning for classification and proposes the One-Pass Multi-View (OPMV) framework which goes through the training data only once without storing the entire training examples. This approach jointly optimizes the composite objective functions with consistency linear constraints for different views. We verify, both theoretically and empirically, the effectiveness of the proposed algorithm. ER -
APA
Zhu, Y., Gao, W. & Zhou, Z.. (2016). One-Pass Multi-View Learning. Asian Conference on Machine Learning, in Proceedings of Machine Learning Research 45:407-422 Available from https://proceedings.mlr.press/v45/Zhu15.html.

Related Material