Bipolar Neural Network Research Articles

Cycles of a discrete time bipolar artificial neural network

A discrete time bipolar neural network depending on two parameters is studied. It is observed that its dynamical behaviors can be classified into six cases. For each case, the long time behaviors can be summarized in terms of fixed points, periodic points, basin of attractions, and related initial distributions. Mathematical reasons are supplied for these observations and applications in cellular automata are illustrated.

Simulation of 2-D Bipolar Hopfield Model Optical Neural Network

Hopfield model with bipolar neural states (-1, 1), has an advantage of output signal to-noise ratio 2 times that of the unipolar neural states (0,1). Bipolar implementation is achieved by biasing the interconnection weights to yield a non-negative memory matrix. This added effect is removed from the output by an addition of distributed background and a dynamic thresholding. The output estimate of the bipolar neural network is derived with all its terms unipolarly expressed. We can use unipolar data to have for a bipolar like neural network. Hence the computing can be performed in a single channel system. Computer simulations have been performed for bipolar 2-D optical neural networks.

Implementation of a bipolar neural network using a single optical channel

This PDF contains the communication on Implementation of a bipolar neural network using a single optical channel.

All-optical bipolar neural network with polarization-modulating neurons

A general all-optical implementation scheme for bipolar neural networks is presented. Bipolar operation is realized by exploiting two orthogonal light polarizations. A three-layer feed-forward bipolar network is demonstrated experimentally by using a liquid-crystal light valve for implementing an array of polarization modulating neurons.