一類離散Hopfield網(wǎng)絡(luò)的吸引特性研究
APPROACHES ON ATTRACTION CHARACTORISTICS OF A KIND OF DISCRETE HOPFIELD NETWORKS
-
摘要: 該文以n維超立方體為大規(guī)模局域連接模型的離散Hopfield網(wǎng)絡(luò)(簡稱An網(wǎng))為研究對象,深入討論了這種網(wǎng)絡(luò)的吸引特性,包括吸引子/吸引環(huán)的結(jié)構(gòu)和吸引半徑等,得出了具有相同結(jié)構(gòu)的吸引子/吸引環(huán)具有相同的吸引特性和吸引半徑,并給出了它們的吸引半徑;進(jìn)一步地,本文將上述討論擴(kuò)展到較為一般的離散Hopfield網(wǎng)絡(luò),討論了它們的吸引子/吸引環(huán)與An網(wǎng)的吸引子/吸引環(huán)相同(或部分相同)的條件,以及相應(yīng)的吸引子/吸引環(huán)的吸引半徑,得出了遷移越大,保留An網(wǎng)吸引子/吸引環(huán)的數(shù)目越少,且吸引半徑越小的結(jié)論。Abstract: This paper studies the attraction characteristics of a kind of symmetric discrete Hopfield neural networks, the network with ri-cube as its structure and connection situation, called n-cube network. The structure of the attraction states/cycles and their attraction diameters are explored. The result is that the attraction region is of same size and so is the attraction diameters for the attraction states/cycles with same attraction structure. Furthermore, the above result is extended to a more general asymmetric discrete Hopfield network and the conditions for the network to be (or partially to be) with the same attraction states/cycles as those of n-cube network are studied, and its attraction diameters are presented.
-
許進(jìn),保錚,超立方體的譜,工程數(shù)學(xué)學(xué)報(bào),1999,16(4),1-5.[2]汪云九,神經(jīng)信息的編碼,97中國神經(jīng)計(jì)算大會(huì)論文集(CCNS97),武漢,1997,Vol.1,32-37[3]張軍英,許進(jìn),保錚,一類離散Hopfield網(wǎng)絡(luò)的吸引狀態(tài)的特征分析,電子與信息學(xué)報(bào),2001,23(7),677 686[4]閻平凡,黃端旭,人工神經(jīng)網(wǎng)絡(luò)-模型,分析與應(yīng)用,合肥安徽教育出版社,1991年,第三章[5]董繼揚(yáng),徐慎初,王新明,均勻分類DHNN的幾何描述,97中國神經(jīng)計(jì)算大會(huì)論文集(CCNS97),武漢,1997.Vol.2.288-292.[6]張軍英,許進(jìn),保錚,Hopfield網(wǎng)絡(luò)的關(guān)聯(lián)分析,自動(dòng)化學(xué)報(bào),1997,23(4),446-454.[7]廖曉昕,昌莉,離散Hopfield神經(jīng)網(wǎng)絡(luò)的穩(wěn)定性研究,自動(dòng)化學(xué)報(bào),1999,25(6),721-727 ［8］F. Luccio, L. Pagli, On a new Boolean function with applications, IEEE Trans. on Computers,1999, 48(3), 296-310.[8]董繼揚(yáng),徐慎初,可用幾何方法描述分類性能的神經(jīng)網(wǎng)絡(luò)的構(gòu)造,神經(jīng)網(wǎng)絡(luò)理論與應(yīng)用研究96,成都,西南交通大學(xué)出版社,1996,129-133.[9]J.J. Hopfield, Neural networks and physical system with emergent collective computer abilities,Proc. Natl. Acad. USA, 1982, 79(4), 2554-2558. -
計(jì)量
- 文章訪問數(shù): 2017
- HTML全文瀏覽量: 136
- PDF下載量: 453
- 被引次數(shù): 0
下載:
