TY - GEN
T1 - Augmenting supervised neural networks with unsupervised objectives for large-scale image classification
AU - Zhang, Yuting
AU - Lee, Kibok
AU - Lee, Honglak
PY - 2016
Y1 - 2016
N2 - Unsupervised learning and supervised learning are key research topics in deep learning. However, as high-capacity supervised neural networks trained with a large amount of labels have achieved remarkable success in many computer vision tasks, the availability of large-scale labeled images reduced the significance of unsupervised learning. Inspired by the recent trend toward revisiting the importance of unsupervised learning, we investigate joint supervised and unsupervised learning in a large-scale setting by augmenting existing neural networks with decoding pathways for reconstruction. First, we demonstrate that the intermediate activations of pretrained large-scale classification networks preserve almost all the information of input images except a portion of local spatial details. Then, by end-to-end training of the entire augmented architecture with the reconstructive objective, we show improvement of the network performance for supervised tasks. We evaluate several variants of autoencoders, including the recently proposed "what-where" autoencoder that uses the encoder pooling switches, to study the importance of the architecture design. Taking the 16-layer VGGNet trained under the ImageNet ILSVRC 2012 protocol as a strong baseline for image classification, our methods improve the validation-set accuracy by a noticeable margin.
AB - Unsupervised learning and supervised learning are key research topics in deep learning. However, as high-capacity supervised neural networks trained with a large amount of labels have achieved remarkable success in many computer vision tasks, the availability of large-scale labeled images reduced the significance of unsupervised learning. Inspired by the recent trend toward revisiting the importance of unsupervised learning, we investigate joint supervised and unsupervised learning in a large-scale setting by augmenting existing neural networks with decoding pathways for reconstruction. First, we demonstrate that the intermediate activations of pretrained large-scale classification networks preserve almost all the information of input images except a portion of local spatial details. Then, by end-to-end training of the entire augmented architecture with the reconstructive objective, we show improvement of the network performance for supervised tasks. We evaluate several variants of autoencoders, including the recently proposed "what-where" autoencoder that uses the encoder pooling switches, to study the importance of the architecture design. Taking the 16-layer VGGNet trained under the ImageNet ILSVRC 2012 protocol as a strong baseline for image classification, our methods improve the validation-set accuracy by a noticeable margin.
UR - http://www.scopus.com/inward/record.url?scp=84998996019&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84998996019&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:84998996019
T3 - 33rd International Conference on Machine Learning, ICML 2016
SP - 939
EP - 957
BT - 33rd International Conference on Machine Learning, ICML 2016
A2 - Balcan, Maria Florina
A2 - Weinberger, Kilian Q.
PB - International Machine Learning Society (IMLS)
T2 - 33rd International Conference on Machine Learning, ICML 2016
Y2 - 19 June 2016 through 24 June 2016
ER -