Introducing a vertical viral infection method for solving problems with evolutionary programming

• Authors: Jonathan W. Lartigue
• Venue: Proceedings of the 46th Annual Southeast Regional Conference on XX
• Year: 2008

Compared to standard search algorithms, which can become unwieldy as the solution space expands, genetic algorithms can often provide a near-optimal solution in fraction of the time. However, genetic algorithms add additional complications, such as stalling at local maxima or minima in the solution space, that can sometimes return an inferior candidate solution. Many resolutions to these problems have been proposed to introduce variation in candidate solutions in order to provide a nudge to genetic algorithms when they become stuck. An alternative that has gained acceptance in recent years is "viral infection," which provides a simple method of introducing new material into the solution space. In contrast to mutation, which introduces new information vertically, that is, in subsequent generations, viral infection allows the horizontal propagation of new information into the current host population through infection of members of the population by a "virus." The approach to viral infection presented in this paper departs dramatically from previously published implementations by incorporating viral infection into the genetic algorithms vertically, which can produce better candidate solutions in less time. This paper will examine several variations to the alternative viral infection method described and will compare and contrast their results on a simple problem space.