On the Improvement of Multi-nary Content Addressable Memory

Abstract

Aims: Using Simple Artificial Neural Networks, and away from strict Boolean logic, this paper proposes a new design of memory array that has the ability to recognize erroneous and deformed data and specify the rate of error. Methodology: To achieve this work, artificial neural network was exploited to be the actor responsible of representing the crude of the building. It’s worth mentioning that simple neurons with binary step function and identity function were used, which will facilitate the way of implementation. The connection of few neurons in a simple network issues an exclusive X gate, which accepts only one value X (where X ∊ ℝ+) with an acceptable error rate α. This gate will be the main core of designing a memory cell that can learn a value X and recognized this value when requested. Results: After several stages of development, the final version of this memory cell will serve as a node unit of a large memory array which can recognize a data word or even a whole image with the ability to accept and recognize distorted data. Specific software that simulates the designed networks was developed in order to declare the efficiency of this memory. The obtained result will judge the Network.