Tuesday, 25 January 2011: 2:15 PM
2A (Washington State Convention Center)
Evolutionary programming has been available since the 1960s: first in the form of "genetic algorithms" (Holland 1975), and later as "genetic programming" (Cramer 1985, Koza 1992). A third, relatively new, variant is called "gene expression programming", as pioneered by Ferreira (2006). The main advantage of gene expression programming (GEP) over the previous evolutionary methods is that GEP is extremely efficient at creating and testing viable candidate functions. Regardless of the complexity of the mutation, all individuals are always computationally viable. This allows the evolution to converge quickly to a solution, and requires no more computer power than a desktop PC. GEP achieves this efficiency by coding each chromosome via book-reading representation of the expression tree, not as a hierarchical representation. We will explain how GEP works, and will show how it can be applied for: ensemble-average precipitation forecasts, probability forecasts of precipitation, and electrical-load forecasts -- all for a region of very complex mountainous terrain.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner