Soft Computing for Image and Multimedia Data Processing by Siddhartha Bhattacharyya & Ujjwal Maulik

Author:Siddhartha Bhattacharyya & Ujjwal Maulik
Language: eng
Format: epub
Publisher: Springer Berlin Heidelberg, Berlin, Heidelberg

5.2 Bidirectional Self-Organizing Neural Network (BDSONN) Architecture

The bidirectional self-organizing neural network (BDSONN) architecture [88, 241–244], as the name suggests, is a three layer network structure assisted by bidirectional propagation of network states for self-supervised organization of input information. The network architecture consists of an input layer for accepting external world inputs and two competing self-organizing network layers, viz., an intermediate layer and an output layer. The number of neurons in each of the network layers corresponds to the number of pixels in the input image. The input layer of the network accepts fuzzy membership values of the constituent pixels in the input image. This fuzzy input information is propagated to the other network layers for further processing. Thus the network layers resemble fuzzy layers of neurons, guided by fuzzy membership information. The neurons in each layer of the network are connected to each other within the same layer following a cellular network structure. The strengths of these intra-layer interconnections are fixed and full and equal to 1. However, each neuron in a particular layer of the network is connected to the corresponding neuron and to its second-order neighbors of the previous layer following a neighborhood-based topology through forward path inter-layer interconnections. In addition, the output layer neurons are similarly connected to the intermediate layer neurons via the backward path inter-layer interconnection paths. The strengths of these inter-layer interconnections between the input layer and the intermediate layer, between the intermediate layer and the output layer and between the output layer and the intermediate layer neurons are decided by the relative measures of the membership values at the individual neurons of the different layers. Figure 5.1 shows a schematic of the proposed BDSONN architecture [88, 241–244] using fully connected network layers and second-order neighborhood topology-based inter-layer interconnections.

Fig. 5.1Bidirectional self-organizing neural network (BDSONN) architecture using a second-order neighborhood-based forward and backward path inter-layer interconnections for the propagation of network states (bold lines indicate path for propagation of fuzzy context-sensitive thresholding information, not all intra-layer interconnections are shown for clarity)

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.