Author

Ravichandran

Date of Award

12-1998

Degree Name

Master of Science in Engineering

Department

Electrical and Computer Engineering

First Advisor

Dr. Garrison Greenwood

Second Advisor

Dr. Hossein Mousavineshad

Third Advisor

Dr. Damon Miller

Access Setting

Masters Thesis-Open Access

Abstract

The exponentially large size of the solution space of typical optimization problems precludes the use of any deterministic approach to search for an optimal solution. A more efficient and efficacious approach is to use heuristic algorithms based on rules of thumb to guide the search process in the solution space. Associated with every solution in the space is a real number called fitness that signifies the quality of a solution. This space and the fitness values together form the fitness landscapes. Knowledge about the topology of these fitness landscapes is vital for any heuristic search operator to expedite the search process and also, find a good solution.

Complete enumeration of the landscape is usually impractical. Moreover, the landscape is often n-dimensional with n >> 3, making it difficult to visualize. This research deals with the development of mathematical and graphical techniques to characterize the structure of fitness landscapes. A new 3-D graphical tool that can depict the topology of high dimensional fitness landscapes has been developed. This graphical approach provides a visual perception of the space to be explored, which can be used to guide and accelerate the search process.

Share

COinS