1: \begin{abstract}

2: With the rapid development of information technologies, centralized

3: data processing is subject to many limitations, such as computational

4: overheads, communication delays, and data privacy leakage. Decentralized

5: data processing over networked terminal nodes becomes an important

6: technology in the era of big data. Dictionary learning is a powerful

7: representation learning method to exploit the low-dimensional structure

8: from the high-dimensional data. By exploiting the low-dimensional

9: structure, the storage and the processing overhead of data can be

10: effectively reduced. In this paper, we propose a novel decentralized

11: complete dictionary learning algorithm, which is based on $\ell^{4}$-norm

12: maximization. Compared with existing decentralized dictionary learning

13: algorithms, comprehensive numerical experiments show that the novel

14: algorithm has significant advantages in terms of per-iteration computational

15: complexity, communication cost, and convergence rate in many scenarios.

16: Moreover, a rigorous theoretical analysis shows that the dictionaries

17: learned by the proposed algorithm can converge to the one learned

18: by a centralized dictionary learning algorithm at a linear rate with

19: high probability under certain conditions.

20: \end{abstract}