Improving dictionary learning using the Itakura-Saito divergence
Improving dictionary learning using the Itakura-Saito divergence
复制标题
DOI：
10.1109/chinasip.2014.6889341
复制
发表时间：
2014-07
期刊：
影响因子：
--
通讯作者：
Zhenni Li;Shuxue Ding;Yujie Li;Zunyi Tang;Wuhui Chen
中科院分区：
文献类型：
--
作者：
Zhenni Li;Shuxue Ding;Yujie Li;Zunyi Tang;Wuhui Chen
针对信号的非负稀疏表示(NNSR)问题，提出了一种改进的、高效的超完备非负字典学习算法。我们采用Itakura-Saito(IS)散度作为误差度量，这与传统的使用欧几里得(EUC)距离作为误差度量的词典学习方法有很大不同。此外，为了加强系数矩阵的稀疏性，我们采用了ℓ1-范数最小化作为稀疏约束。词典恢复的数值实验表明，本文提出的词典学习算法比其他现有的以欧氏距离作为误差度量的算法具有更好的性能。
This paper presents an improved and efficient algorithm for overcomplete, nonnegative dictionary learning for nonnegative sparse representation (NNSR) of signals. We adopt the Itakura-Saito (IS) divergence as the error measure, which is quite different from the conventional dictionary learning methods using the Euclidean (EUC) distance as the error measure. In addition, for enforcing the sparseness of coefficient matrix, we impose ℓ1-norm minimization as the sparsity constraint. Numerical experiments on recovery of a dictionary show that the proposed dictionary learning algorithm performs better than other currently available algorithms which use Euclidean distance as the error measure.