A VLSI BAM neural network chip for pattern recognition applications

Abstract

ABSTAACTNeural networks are presently being extensively explored for use in Pattern Recognition applications. Bi-directional Associative Memory (BAM) is a two-level non-linear neural network suitable for pattern recognition applications. One important performance attribute of the discrete BAM is its ability to recall stored pattern pairs, particularly in the presence of noise. In this paper the VLSI implementation of BAM is presented. A modular VLSI processor chip implementing BAM was designed. By using 2 micron CMOS technology, 4 neurons with 8 modules of 256x5 bit local weightstorage memory were integrated on a 6.9 x 7.4 mm2 die. With 4 operating modes (Learn, Evaluate, Read and Write), it is suitable to serve as a eo-processor. The system architecture is highlyflexible and modular, enabling the construction of larger BAM networks of upto 252 neurons using multiple BAM chips. Results show that real-time speeds can be achieved. The total training time for a full network of upto 252 neurons is 1.5997 ms at a clock-rate of lOh4HZ, which is fast eneough for numerous pattern recognition applications.

Novel stability criteria for bidirectional associative memory neural networks with time delays

“…Now, calculate and obtain 1 = 2 =2:18¿2; 1 = 2 =2:18¿2; Á 1 =Á 2 =2:3884¿2 if we select i = 2+j =1 (i; j =1; 2). Hence, conditions (9), (17), and (21) given in this paper are not satisÿed. For numerical simulation, system (26) T , and the global stability of system (26).…”

“…Note that i and n+j might indicate some sort of stability margins. (iv) Compute 1 and 2 on the left-hand side of Equation (9). (v) Check whether inequality (9) is satisÿed; if not, return to Step (iii) and increase the values of i and n+j .…”

“…[1][2][3][4][5][6][7][8][9] and the references cited therein). BAM neural network is described by the following set of di erential equations [1,2,5,10]:…”

A Survey of Neuromorphic Computing and Neural Networks in Hardware

“…Full custom or application specific integrated circuit (ASIC) chips have also been very common for neuromorphic implementations [6], [9], [348], [350], [389]- [404], [438], [477], [677], [683], [749]- [764], [764]- [808], [1047], [1048], [1054]- [1060], [1100], [1101], [1118]- [1130], [1170], [1183]- [1185], [1204], [1205], [1212], [1236]- [1242], [1260], [1275], [1283], [1363], [1410], [1527]- [1561]. IBM's TrueNorth, one of the most popular present-day neuromorphic implementations, is a full custom ASIC design [1562]- [1569].…”

Stochastic Synchronization of Impulsive Reaction–Diffusion BAM Neural Networks at a Fixed and Predetermined Time

“…Bidirectional associative memory (BAM) neural networks are a type of artificial neural network model that can be used to recognize and classify patterns in input data. They can be applied to tasks such as pattern recognition, image classification, speech recognition, and signal processing [1][2][3][4][5]. In BAM networks, information can be stored and retrieved bidirectionally, meaning that patterns can be associated in both forward and backward directions [1].…”