Associative Memory For Pattern Recognition Research Articles

Learning algorithms for oscillatory neural networks as associative memory for pattern recognition

Alternative paradigms to the von Neumann computing scheme are currently arousing huge interest. Oscillatory neural networks (ONNs) using emerging phase-change materials like VO2 constitute an energy-efficient, massively parallel, brain-inspired, in-memory computing approach. The encoding of information in the phase pattern of frequency-locked, weakly coupled oscillators makes it possible to exploit their rich non-linear dynamics and their synchronization phenomena for computing. A single fully connected ONN layer can implement an auto-associative memory comparable to that of a Hopfield network, hence Hebbian learning rule is the most widely adopted method for configuring ONNs for such applications, despite its well-known limitations. An extensive amount of literature is available about learning in Hopfield networks, with information regarding many different learning algorithms that perform better than the Hebbian rule. However, not all of these algorithms are useful for ONN training due to the constraints imposed by their physical implementation. This paper evaluates different learning methods with respect to their suitability for ONNs. It proposes a new approach, which is compared against previous works. The proposed method has been shown to produce competitive results in terms of pattern recognition accuracy with reduced precision in synaptic weights, and to be suitable for online learning.

A prototype of programmable associative memory for track finding

We present a device, based on the concept of associative memory for pattern recognition, dedicated to on-line track finding in high-energy physics experiments. A large pattern bank, describing all possible tracks, can be organized into field programmable gate arrays where all patterns are compared in parallel to data coming from the detector during readout. Patterns, recognized among 2/sup 66/ possible combinations, are output in a few 30 MHz clock cycles. Programmability results in a flexible, simple architecture and it allows one to keep up smoothly with technology improvements. A 64 PAM array has been assembled on a prototype VME board and fully tested up to 30 MHz.

A programmable associative memory for track finding

We present a device, based on the concept of associative memory for pattern recognition, dedicated to on-line track finding in high-energy physics experiments. A large pattern bank, describing all possible tracks, can be organized into Field Programmable Gate Arrays where all patterns are compared in parallel to data coming from the detector during readout. Patterns, recognized among 2 66 possible combinations, are output in a few 30 MHz clock cycles. Programmability results in a flexible, simple architecture and it allows to keep up smoothly with technology improvements.

Terminal attractor optical associative memory with adaptive control parameter

The control parameter dependence in terminal attractor (TA) optical associative memory is investigated. With adaptive setting of the control parameter in the dynamics of the TA model for dynamic iteration, perfect convergence of the associative memory for pattern recognition is achieved. With numerical simulations, the optimal control parameter in the TA model associative memory is determined. The optimal control parameter is also used in an optical experiment. The experimental results show that the target pattern can be successfully associated.

Terminal attractor optical associative memory for pattern recognition

Terminal attractor optical associative memory for pattern recognition

The AMchip: a full-custom CMOS VLSI associative memory for pattern recognition

An associative memory full-custom CMOS VLSI chip (AMchip), to be used in fast trigger systems for pattern recognition, has been designed and is being tested. The AMchip is a full-custom associative memory IC developed for high energy physics. It contains about 140000 MOS transistors, has been realized in 1.5- mu m, double-metal, silicon gate CMOS technology, and is housed in a 120-pin package. The AMchip has been designed to be used with any kind of detector which provides in the output the hit coordinates. Some preliminary tests have been done on the 55 prototypes received from the ES2 (European silicon structures) foundry. The performance test, limited by the equipment used, has shown that the AMchip can work with a 20-MHz clock.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Evaluation and improvement of associative memory for pattern recognition

Optical systems of associative memory models are capable of reconstructing lost information from incomplete input patterns. In this paper, we evaluate the ability of associative memory models when used in practical pattern recognition systems. Effects of improvement techniques, including defocusing preprocessing, and the recursive use of the associative operation are also examined with computer simulations. Finally the effectiveness of the recognition system is tested experimentally with alphabetic characters in different illumination conditions.