Photo credit: Photo by John Karlo Mendoza on Unsplash.

Editors note: Dr. Yampolskiy is Associate Professor in the department of Computer Engineering and Computer Science at the Speed School of Engineering, University of Louisville. In this series, he asks: What Can and Cant Darwins Algorithm Compute? See also yesterdays post, the first in the series, What Can and Cant Darwins Algorithm Compute?

Inspired by Darwins theory1 of biological evolution, evolutionary computation attempts to automate the process of optimization and problem solving by simulating differential survival and reproduction of individual solutions. From the early 1950s, multiple well-documented attempts to make Darwins algorithm work on a computer have been published under such names as Evolutionary Programming12, Evolutionary Strategies13, Genetic Algorithms14, Genetic Programming15, Genetic Improvement16, Gene Expression Programming17, Differential Evolution18, Neuroevolution19, and Artificial Embryogeny20. While numerous variants different in their problem representation and metaheuristics exist21-24, all can be reduced to just two main approaches Genetic Algorithm (GA) and Genetic Programming (GP).

GAs are used to evolve optimized solutions to a particular instance of a problem such as Shortest Total Path25, Maximum Clique26, Battleship27, Sudoku28, Mastermind23, Light Up29, Graph Coloring30, integer factorization31, 32, or efficient halftone patterns for printers33, and so are not the primary focus of this paper. GPs purpose, from their inception, was to automate programming by evolving an algorithm or a program for solving a particular class of problems, for example an efficient34 search algorithm. Software design is the type of application most frequently associated with GPs35, but work in automated programming is also sometimes referred to as real programing, object-oriented GP, algorithmic programming, program synthesis, traditional programming, Turing Equivalent (TE) programming or Turing-complete GP36-38.

The sub-field of computation, inspired by evolution in general, and the Genetic Programing paradigm, established by John Koza in 1990s, in particular are thriving and growing exponentially. This is evidenced both by the number of practitioners and of scientific publications. Petke et al. observe enormous expansion of number of publications with the Genetic Programming Bibliography passing 10,000 entries By 2016 there were nineteen GP books including several intended for students 16. Such tremendous growth has been fueled, since the early days, by belief in the capabilities of evolutionary algorithms, and our ability to overcome obstacles of limited computational power or data as illustrated by the following comments:

Tomorrow, State-of-the-Art in Evolutionary Computation.

Here is the original post:

Evolution On Evolutionary Computation - Discovery Institute