I= APPLICATION OF SPATIAL FILTERS TO OPTICAL COMPUTING AND NEURAL NETWORKS
Pakistan Research Repository Home
 

Title of Thesis
APPLICATION OF SPATIAL FILTERS TO OPTICAL COMPUTING AND NEURAL NETWORKS

Author(s)
S. Aftab Ahsan Rizvi
Institute/University/Department Details
Department of Physics/ Quaid-i-Azam University, Islamabad
Session
1993
Subject
Number of Pages
155
Keywords (Extracted from title, table of contents and abstract of thesis)
spatial filters, optical computing, neural networks, neural network model, associative memory, pattern recognition, optical computer, holographic techniques, ternary logic, carry free algorithm, optical binary logic gates, modified signed digit algorithm

Abstract
Digital optical computing has gained a lot of interest because of its fascinating features such as highly parallel processing capability of data at ultra high speed. In this thesis, spatial filtering has been employed to design and experimentally implement the basic components of a proposed digital optical computer.

An associative memory model, i.e., neural network model and its optical implementation using simple grating structures has been presented. The associative memory (filter) has been fabricated using optical and computer generated holographic techniques. These systems have the great potential towards solving fuzzy problems like pattern recognition in an optical computer.

Moreover, application of spatial filtering technique for designing location addressable memory devices based on multiple-valued- logic has also been presented. The schemes have been experimentally implemented using grating structures. It is shown that the ternary logic is found more efficient when compared to the binary logic as it requires less number of digits to process a number. However, the schemes based on multiple-valued logic algorithms have limited parallelism because of the carry/borrow for each input data. A scheme for the optical processor based on carry free algorithm, i.e., modified signed digit using spatial filtering technique has also been presented. Two encoding procedures have been adopted which are polarization and grating structures.

Download Full Thesis
11765.65 KB
S. No. Chapter Title of the Chapters Page Size (KB)
1 0 Contents
826.3 KB
2 1 Introduction 1
1514.08 KB
  1.1 Architecture Of An Optical Computer 4
  1.2 Spatial Filters And Nonolinear Filtering 11
  1.3 Layout Of The Thesis 15
3 2 Optical Pattern Recognition Using Neural Network Models 17
1978.64 KB
  2.1 Hopfield Model 19
  2.2 Associatiy Memory Model/Formalism 22
  2.3 Optical Implementation Using Structures Grating 26
  2.4 Fabrication Of Filter (Associative Memory ) 30
  2.5 Comparison Of The Proposed Optical Implementation Scheme With Other Schemes 34
  2.6 Conclusions 38
4 3 Design Of Optical Processing Elements 40
1939.85 KB
  3.1 Truth Tables For Multiple-Valued Logic 41
  3.2 Optical Binary Logic Gates And Their Combinational Units 45
  3.3 Optical Design Of Ternary Logic Gates, Half-Adder And €“Subtractor 58
  3.4 Conclusions 68
5 4 Optical Computing Based On Modified Signed Digit Algorithm 71
1678.11 KB
  4.1 Modified Signed Digit Number System 74
  4.2 MSD Logic Elements And Arithmetic Operations 76
  4.3 Optical Implementation Of MSD Logic Elements Using Polarization Encoding 85
  4.4 Optical Implementation Of The MSD Logic Elements Using Frequency Modulated Grating Structures 89
  4.5 Optical Processor Using MSD Algorithm Based On Spatial Filtering Technique 91
  4.6 Conclusions 93
6 5 Appendixes 95
4008.42 KB
  5.1 References 133