Genetic Programming (GP) means to perform evolutionary search in a tree-based search space. Often, such trees are designed to represent of sorts, sometimes classifiers. This field is somehow in decline with the advent of deep learning. However, there are quite a few interesting things that can be done.

Algorithm Synthesis

For me, the most interesting topic of Genetic Programming was always . Here, the goal is to use optimization algorithms to construct new algorithms. This can well be understood as a field of , but we tackled it before the advent of code generation. The goal was to find algorithms that produce specific outputs for specific inputs, without any additional description. And it worked quite well and it became my PhD thesis. Matter of fact, combining GP with yields very good performance when constructing discrete algorithms. Sadly, I never had time to follow this line of research more in depth after receiving my PhD, but I hope to do so in the future.

Classification

Classification is another classical application area of Genetic Programming. Since many traditional classification approaches use decision trees in one way or another and GP means searching the space of trees, this comes very natural.

Other Topics

Besides algorithm synthesis and classification, I also applied Genetic Programming to some other domains, such as .

  • Jin OUYANG (欧阳晋), Thomas Weise (汤卫思), Alexandre Devert, and Raymond Chiong: SDGP: A Developmental Approach for Traveling Salesman Problems. IEEE Symposium on Computational Intelligence in Production and Logistics Systems (CIPLS'2013), part of the IEEE Symposium Series on Computational Intelligence (SSCI'2013), April 15-19, 2013, Singapore, pages 78–85, Los Alamitos, CA, USA: IEEE Computer Society Press.
  • Thomas Weise (汤卫思), Alexandre Devert, and Ke TANG (唐珂): A Developmental Solution to (Dynamic) Capacitated Arc Routing Problems using Genetic Programming. 14th Genetic and Evolutionary Computation Conference (GECCO'2012), July 7-11, 2012, Philadelphia, PA, USA, pages 831–838. New York, NY, USA: ACM.
  • Thomas Weise (汤卫思): Global Optimization Algorithms — Theory and Application. self-published, free e-book. 2009.