一種提高遞歸相關(guān)聯(lián)想記憶器(ROAM)容量的方法
AN APPROACH TO INCREASE THE CAPACITY OF RECURRENT CORRELATION ASSOCIATIVE MEMORY
-
摘要: 本文從信噪比分析的角度對遞歸相關(guān)聯(lián)想記憶(RCAM)模型中的非線性函數(shù)如何影響其性能的問題進(jìn)行了研究,給出了一新的非線性函數(shù)。理論分析及仿真實(shí)驗(yàn)皆表明本文的模型有著更好的記憶性能。Abstract: This paper addresses the effect of nonlinearity of RCAM on its recall performance with the help of signal-to-noise analysis and presents a new nonlinear function. Theoretical analysis and computer simulations show that the presented memory model offers more satisfactory performance.
-
Hopfield J J. Neural networks and physical systems with emergent collective computation abilities[J].Proc. Nat. Acad. Sci. USA.1982, 79(7):2554-2558[2]Chiueh T D, Goodman R M. Recurrent correlation associative memories[J].IEEE Trans. on Neural Networks.1991, 2(2):275-284[3]Marks II R J, Atlas L E, Choi J J, et al. Performance analysis of associative memories with poplin-earities in the corelation domain[J].Appl. Opt.1988, 27(14):2900-2904[4]Selviah D R, Midwinter J E, Rivers A W, et al. Correlating matched-filter model for analysis and[5]optimization of neural networks. IEE Proc.-F, 1989, 146(3): 143-148.[6]Psaltis D, Park C H. Hong J. Higher order associative memories and their optical implementation. Neural Networks, 1988, 1(1): 149-163.[7]Wang C C, Don H S. An analysis of high-capacity discrete exponential BAM[J].IEEE Trans. on Neural Networks.1995, 6(2):492-496 -
計(jì)量
- 文章訪問數(shù): 2173
- HTML全文瀏覽量: 132
- PDF下載量: 424
- 被引次數(shù): 0
下載:
