14 May 2010 Fixed weight Hopfield Neural Network based on optical implementation of all-optical MZI-XNOR logic gate
Author Affiliations +
Abstract
Many researches are conducted to improve Hopfield Neural Network (HNN) performance especially for speed and memory capacity in different approaches. However, there is still a significant scope of developing HNN using Optical Logic Gates. We propose here a new model of HNN based on all-optical XNOR logic gates for real time color image recognition. Firstly, we improved HNN toward optimum learning and converging operations. We considered each unipolar image as a set of small blocks of 3-pixels as vectors for HNN. This enables to save large number of images in the net with best reaching into global minima, and because there are only eight fixed states of weights so that only single iteration performed to construct a vector with stable state at minimum energy. HNN is useless in dealing with data not in bipolar representation. Therefore, HNN failed to work with color images. In RGB bands each represents different values of brightness, for d-bit RGB image it is simply consists of d-layers of unipolar. Each layer is as a single unipolar image for HNN. In addition, the weight matrices with stability of unity at the diagonal perform clear converging in comparison with no self-connecting architecture. Synchronously, each matrix-matrix multiplication operation would run optically in the second part, since we propose an array of all-optical XOR gates, which uses Mach-Zehnder Interferometer (MZI) for neurons setup and a controlling system to distribute timely signals with inverting to achieve XNOR function. The primary operation and simulation of the proposal HNN is demonstrated.
© (2010) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Kussay Nugamesh Mutter, Kussay Nugamesh Mutter, Mohd Zubir Mat Jafri, Mohd Zubir Mat Jafri, Azlan Abdul Aziz, Azlan Abdul Aziz, } "Fixed weight Hopfield Neural Network based on optical implementation of all-optical MZI-XNOR logic gate", Proc. SPIE 7717, Optical Modelling and Design, 77170X (14 May 2010); doi: 10.1117/12.854838; https://doi.org/10.1117/12.854838
PROCEEDINGS
8 PAGES


SHARE
RELATED CONTENT

Neural Networks For Logic Operations
Proceedings of SPIE (October 24 1989)
Binary neuron with optical devices
Proceedings of SPIE (January 06 2009)
Mass flow rate measurement of oil water two phase flow...
Proceedings of SPIE (October 12 2008)
Neural network optimization, components, and design selection
Proceedings of SPIE (December 31 1990)

Back to Top