Illinois Genetic Algorithms Laboratory

One of the key research areas in IlliGAL is linkage learning. Many of the GAs with linkage learning ability (e.g. LLGA, eCGA, BOA, DSMGA) developed in IlliGAL perform on binary of k-ary domain. To be able to perform those linkage learning techniques on real-value domain, there are at least two possible ways to go. One is to encode real values into a binary string. In one of the paper that I worked with Scott G. Santarelli and David E. Goldberg, we used such method and applied hBOA to optimize an antenna array. The results indicate that linkage learning techniques do help when dealing with difficult problems. However, this encoding-real-into-binary scheme wouldn’t tell us directly which parameters are highly depend, and also the encoding might increase the difficuly of linkage learning. The other way to go is to directly extend those linage learning techniques to real-value domain. Chang Wook Ahn’s thesis reveals a way of doing so, in which he extended hBOA into real-valued domain. With the growing need of GA in real world, we need to investigate more in this area.