Sparse Distributed Memory with Monotonic Decision Function


The KIPS Transactions:PartB , Vol. 8, No. 1, pp. 105-113, Feb. 2001
10.3745/KIPSTB.2001.8.1.105,   PDF Download:

Abstract

Sparse Distributed Memory (SDM) has been proposed as a practical neural network model due to its adaptability in problem solving and simplicity in hardware implementation. SDM, however, becomes an inefficient model in case there are inequality relations among the objects in the solution space like real number, because the neurons of a SDM bisect a solution space into the inside and the outside of a circle with a specific radius, and are unioned simply. On the other hand, the neurons of a multilayer perceptron (MLP) bisect a solution space using a linear/nonlinear decision function, and are variously combined with each other to solve general problems. Thus it can solve a general problem. In this paper, we study the characteristics and the cause of the inefficiency of a SDM, and propose a modified SDM which can solve it effectively when a solution space is divided into two regions by a monotonic decision function. To solve it, we introduce a magnitude comparing step into the conventional SDM algorithm. In addition, we show the experimental result by applying the proposed model to an ATM call admission control.


Statistics
Show / Hide Statistics

Statistics (Cumulative Counts from September 1st, 2017)
Multiple requests among the same browser session are counted as one view.
If you mouse over a chart, the values of data points will be shown.


Cite this article
[IEEE Style]
H. Y. Kwon, J. W. Jang, S. J. Lim, D. S. Cho, H. Y. Hwang, "Sparse Distributed Memory with Monotonic Decision Function," The KIPS Transactions:PartB , vol. 8, no. 1, pp. 105-113, 2001. DOI: 10.3745/KIPSTB.2001.8.1.105.

[ACM Style]
Hee Yong Kwon, Jung Woo Jang, Sung Joon Lim, Dong Sub Cho, and Hee Yeung Hwang. 2001. Sparse Distributed Memory with Monotonic Decision Function. The KIPS Transactions:PartB , 8, 1, (2001), 105-113. DOI: 10.3745/KIPSTB.2001.8.1.105.