| Abstract: | Generalisation is one of the most important performance evaluationcriteria for artificial learning systems. An increasing amount ofresearch has recently concentrated on the robustness or generalisationability of the programs evolved using Genetic Programming (GP). Whilesome of these researchers report on the brittleness of the solutionsevolved, some others propose methods of promotingrobustness/generalisation. In this paper, a review of research ongeneralisation in GP and problems with brittleness of solutions producedby GP is presented. Also, a brief overview of several new methodspromoting robustness/generalisation of the solutions produced by GP arepresented. |
