Date of Submission


Date of Award


Institute Name (Publisher)

Indian Statistical Institute

Document Type

Doctoral Thesis

Degree Name

Doctor of Philosophy

Subject Name

Computer Science


Machine Intelligence Unit (MIU-Kolkata)


Pal, Sankar Kumar (MIU-Kolkata; ISI)

Abstract (Summary of the Work)

Pattern recognition and machine learning form a major area of research and develop- ment activity that encompasses the processing of pictorial and other non-numerical information obtained from the interaction between science, technology and society. A motivation for the spurt of activity in this field is the need for people to com- municate with the computing machines in their natural mode of communication. Another important motivation is that the scientists are also concerned with the idea of designing and making intelligent machines that can carry out certain tasks that we human beings do. The most salient outcome of these is the concept of future generation computing systems.Machine recognition of patterns can be viewed as a two fold task, comprising learning the invariant properties of a set of samples characterizing a class, and of deciding that a new sample is a possible member of the class by noting that it has properties common to those of the set of samples. The tasks required for developing and imple- menting a decision rule can be described as a transformation from the measurement space M to the feature space F and finally to the decision space D, i.e.,M ⟶ F ⟶ D.Here the mapping 6: F⟶ D is the decision function, and the elements de D are termed as decisions. Genetic algorithms (GAs) [22, 35, 46, 51, 58, 61, 63, 64, 73, 81, 89, 126, 17o, 206] are randomized search and optimization techniques guided by the principles of evolution and natural genetics. The term was first mentioned by Bagley in 1967 [8], when he devised a genetic algorithm based game playing program using some commonly used operators. He found that the GA was insensitive to the game non-linearity, and performed well over a range of environments. It was then with the pioneering work of Holland in 1975 (89] that GAs were firmly established as an effective search and optimization strategy. GAs mimic some of the processes observed in natural evolution, which include op- erations like selection, crossover and mutation. They perform multimodal search in complex landscapes and provide near optimal solutions for objective or fitness function of an optimization problem. They are efficient, adaptive and robust search processes, with a large amount of implicit parallelism [73, 89]. Genetic algorithms are a relatively recent development, and are gradually finding widespread applications in solving problems requiring efficient and effective search, in business, scientific and engineering circles [22, 58, 64, 68, 81, 153, 170].Many tasks involved in the process of recognizing a pattern need appropriate pa- rameter selection and efficient search in complex and large spaces in order to attain optimal solutions. This makes the process not only computationally intensive, but also leads to a possibility of losing the exact solution. Therefore, the application of GAs for solving certain problems of pattern recognition, that require optimization of computation requirements, and robust, fast and close approximate solution, seems appropriate and natural. Additionally, the existence of the proof of convergence of GAs to the global optimal solution as the number of iterations goes to infinity [28], further strengthens the theoretical basis of its use in search problems.


ProQuest Collection ID:

Control Number


Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.


Included in

Mathematics Commons