Blockwise coordinate descent schemes for efficient and effective dictionary learning
详细信息 查看全文
- 作者：Bao-Di Liua ; thu.liubaodi@gmail.com" class="auth_mail" title="E-mail the corresponding authorAuthor Vitae ; Yu-Xiong Wangb ; yuxiongw@cs.cmu.edu" class="auth_mail" title="E-mail the corresponding authorAuthor Vitae ; Bin Shenc ; stanshenbin@gmail.com" class="auth_mail" title="E-mail the corresponding authorAuthor Vitae ; Xue Lid ; lixue421@gmail.com" class="auth_mail" title="E-mail the corresponding authorAuthor Vitae ; Yu-Jin Zhangd ; zhang-yj@mail.tsinghua.edu.cn" class="auth_mail" title="E-mail the corresponding authorAuthor Vitae ; Yan-Jiang Wanga ; yjwang@upc.edu.cn" class="auth_mail" title="E-mail the corresponding authorAuthor Vitae
- 关键词：Dictionary learning ; Coordinate descent ; Sparse representation ; Image classification
- 刊名：Neurocomputing
- 出版年：2016
- 出版时间：20 February 2016
- 年：2016
- 卷：178
- 期：Complete
- 页码：25-35
- 全文大小：2690 K
文摘
- class="listitem" id="list_li0005">
- class="label">•
A novel simple and efficient algorithm BCDDL is proposed to solve SC–DL problems.
- class="label">•
BCDDL is the fastest algorithm in solving SC–DL to date.
- class="label">•
BCDDL shows state-of-the-art performance when learning bases with small samples.
- class="label">•
BCDDL shows state-of-the-art performance when pursuing comparatively much sparser codes.
- class="label">•
BCDDL achieves superior performance in image classification tasks.