Deep Boosting: Joint feature selection and analysis dictionary learning in hierarchy

NEUROCOMPUTING

卷:178

Article

Peng, Zhanglin¹  Li, Ya²  Cai, Zhaoquan⁴  Lin, Liang³ 

[1] Sun Yat Sen Univ, Guangzhou 510275, Guangdong, Peoples R China

[2] Sun Yat Sen Univ, Sch Informat Sci & Technol, Guangzhou 510275, Guangdong, Peoples R China

[3] Sun Yat Sen Univ, Sch Data & Comp Sci, Guangzhou 510275, Guangdong, Peoples R China

[4] Huizhou Univ, Huizhou, Peoples R China

关键词: Representation Learning;    Compositional boosting;    Dictionary learning;    Image Classification;   

DOI  :  10.1016/j.neucom.2015.07.116

来源: Elsevier

【 摘 要 】

This work investigates how the traditional image classification pipelines can be extended into a deep architecture, inspired by recent successes of deep neural networks. We propose a deep boosting framework based on layer-by-layer joint feature boosting and dictionary learning. In each layer, we construct a dictionary of filters by combining the filters from the lower layer, and iteratively optimize the image representation with a joint discriminative-generative formulation, i.e. minimization of empirical classification error plus regularization of analysis image generation over training images. For optimization, we perform two iterating steps: (i) to minimize the classification error, select the most discriminative features using the gentle adaboost algorithm; (ii) according to the feature selection, update the filters to minimize the regularization on analysis image representation using the gradient descent method. Once the optimization is converged, we learn the higher layer representation in the same way. Our model delivers several distinct advantages. First, our layer-wise optimization provides the potential to build very deep architectures. Second, the generated image representation is compact and meaningful by jointly considering image classification and generation. In several visual recognition tasks, our framework outperforms existing state-of-the-art approaches. (C) 2015 Elsevier B.V. All rights reserved.

【 授权许可】

Free