|
Research |
Over several decades of use in diverse scientific and engineering fields, evolutionary optimization has acquired a reputation for being a kind of universal acid—a general purpose approach that routinely procures useful solutions to optimization problems with rugged, dynamic, and stochastic cost functions over search spaces consisting of strings, vectors, trees, and instances of other kinds of data structures. Remarkably, despite years of research, the means by which evolutionary algorithms work is not well understood.
I've
developed a new explanation for adaptation in genetic algorithms. This account—
the
generative fixation hypothesis—is
based on evidence that genetic algorithms can implement a stochastic
non-local search heuristic, called hyperclimbing,
extraordinarily efficiently. The generative fixation
hypothesis departs from the reigning hypothesis in genetic algorithmics—the building block hypothesis—at a
fundamental level. The potential impact of this new hypothesis extends
to
the fields of evolutionary computation, optimization, machine learning,
theoretical computer science, and evolutionary biology.
|
|
Publications & Manuscripts |
| |
- Explaining Adaptation in Genetic Algorithms with
Uniform Crossover, [pdf]
Keki M. Burjorjee
To appear in Foundations of Genetic Algorithms Conference 2013
- Generative
Fixation: A Unified Explanation for the Adaptive Capacity of
Simple Recombinative Genetic Algorithms
[website]
Keki M. Burjorjee
Ph.D. Thesis, Brandeis University, August 2009
- Sufficient Conditions for
Coarse-Graining Evolutionary Dynamics,
[pdf], slides [pdf]
Keki M. Burjorjee
In Foundations of Genetic Algorithms
Conference IX, 2007
- A
General Coarse-Graining Framework for Studying Simultaneous
Inter-Population Constraints Induced by Evolutionary Operations, [pdf],
extended version with proofs [pdf], slides
[pdf]
Keki M. Burjorjee, Jordan B. Pollack
In
Proceedings of the 2006 Genetic and Evolutionary Computation
Conference
- Theme
Preservation and the Evolution of Representation,
[pdf]
Keki M. Burjorjee, Jordan B. Pollack
In Proceedings of
the 2nd Indian
International Conference on Artificial Intelligence, 2005
- Theme
Preservation and the Evolution of Representation,
[pdf]
Keki M. Burjorjee, Jordan B. Pollack
In Proceedings of
the Theory of
Representation Workshop at the Genetics and Evolutionary Computation
Conference, 2005
|
|
Awards & Honors
|
| |
- Winner of the Tiny
GA competition (GECCO 2006)
- Awarded the Sproull Fellowship for "unusually strong
potential for graduate study" (University of Rochester, 2000)
- Awarded the Mary Evelyn Wells and Gertrude Smith
Prize for excellence in the study of undergraduate Mathematics (Vassar
College, 1998)
- General honors, and departmental honors in mathematics, and computer science (Vassar College, 1998)
- Part of a team of
three that placed second in a national high school programming competition (Computer Society of India, 1992)
|
|
Software |
SpeedyGA:
A fast Simple Genetic Algorithm in Matlab.
|
|
|
