Learning Deep Representation for Imbalanced Classification

Chen Huang^1,2, Yining Li¹, Chen Change Loy^1,3, and Xiaoou Tang^1,3

¹Department of Informaiton Engineering, The Chinese University of Hong Kong, ²SenseTime Group Limited,

³Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences.

IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2016, Spotlight Presentation

News

An extended version of this paper is available. Check Deep Imbalanced Learning for Face Recognition and Attribute Prediction, arXiv:1806.00194, 2018.

Abstract

Data in vision domain often exhibit highly-skewed class distribution, i.e., most data belong to a few majority classes, while the minority classes only contain a scarce amount of instances. To mitigate this issue, contemporary classification methods based on deep convolutional neural network (CNN) typically follow classic strategies such as class re-sampling or cost-sensitive training. In this paper, we conduct extensive and systematic experiments to validate the effectiveness of these classic schemes for representation learning on class-imbalanced data. We further demonstrate that more discriminative deep representation can be learned by enforcing a deep network to maintain both inter-cluster and inter-class margins. This tighter constraint effectively reduces the class imbalance inherent in the local data neighborhood. We show that the margins can be easily deployed in standard deep learning framework through quintuplet instance sampling and the associated triple-header hinge loss. The representation learned by our approach, when combined with a simple k-nearest neighbor (kNN) algorithm, shows significant improvements over existing methods on both high- and low-level vision classification tasks that exhibit imbalanced class distribution.

Deep Learning Large Margin Local Embedding (LMLE)

From Class-level Triplet Embedding to Cluster- & Class-level Quintuplet Embedding: preserving both locality across clusters and discrimination between classes, thus forming the local classification boundary resistent to class imbalance

Training by a triple-header hinge loss to realize this objective
Equal class re-sampling & class costs assignment in mini-batches
Testing: Large margin cluster-wise kNN search

Results

Face attribute prediction: highly imbalanced positive vs. negative class samples (average positive class rate 23% for 40 attributes)
Edge detection: long-tailed distribution of edge structures

Citation

@inproceedings{huang2016lmle,
 author = {Chen Huang and Yining Li and Chen Change Loy and Xiaoou Tang},
 title = {Learning Deep Representation for Imbalanced Classification},
 booktitle = {IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
 year = {2016},}