The pupose of this project is to collect diffrent fitness landscape analysis metrics and diagnostic problem sets that allow practitioners to test and analyze different evolutionary computation algorithms. Problem sets vary in solution representation, some of which include combinitorial, numberical, or algorithmic solutions. Fitness landscape analysis meterics capture landscape features and algorithm characteristics. Below are some diagnostic/benchmark problem sets and fitness landscape metrics that practitioners can use to evaluate evolutionary compuation algorithms.
- Problem Characteristics Diagnostics
- NK Landscapes (and variants)
- Program Synthesis Benchmark Suite
- Function Optimization Problems (GECCO Niching Competions)
- Input-output Machine Learning (Dow Chemical, others)
- Autocorrelation Coefficient
- Fitness Distance Correlation
- Fitness Clouds Plots
- Fitness Density Plots
- Random Walk Correlation