用于數(shù)字式細(xì)胞神經(jīng)網(wǎng)絡(luò)設(shè)計的學(xué)習(xí)算法
LEARNING ALGORITHM FOR THE DESIGN OF DIGITAL CELLULAR NEURAL NETWORK
-
摘要: 本文針對數(shù)字式細(xì)胞神經(jīng)網(wǎng)絡(luò)(DCNN)的應(yīng)用,給出了DCNN模板的設(shè)計方法,即不等式構(gòu)造法,并提出一個基于松弛法的DCNN有教師學(xué)習(xí)算法,為DCNN的設(shè)計提供了理論根據(jù)。在連通片檢測等應(yīng)用中的模擬表明了算法的有效性和正確性。Abstract: The application of cellular neural networks is determined by its templates. In accordance with the application of digital cellular neural network the author proposed before, a method for the design of its templates, i. e., inequalities construcion method, is given and a supervised learning algorithm is proposed based on the relaxation method. The learning algorithm provides theoretical basis for the design of DCNN. Simulation results on examples such as connected component detection shows the effectiveness and feasibility of the algorithm.
-
Chua L O, Yang L. Cellular neural networks:Theory and applications. IEEE Trans. on CAS, 1988, 35(10): 1257-1290.[2]Chua L O, Thiran P. An Analytic method for designing simple CNNs. IEEE Trans. on CAS, 1991, 38(11): 1332-1341.[3]喬長閣.數(shù)字式細(xì)胞神經(jīng)網(wǎng)絡(luò)理論與應(yīng)用研究:[博士論文].西安:西北工業(yè)大學(xué),1993.[4]Matsumoto T, et al. CNN cloning template:Connected component detector. IEEE Trans. on CAS, 1990, 37(5):633-635.[5]Motzkin T S, Schoenberg L J. The Relaxation method for linear inequalities[J].Canada Journal of Mathematics.1954, 6(3):393-404 -
計量
- 文章訪問數(shù): 2093
- HTML全文瀏覽量: 100
- PDF下載量: 348
- 被引次數(shù): 0
下載:
