Tag-accessed memory for genetic programming

Here, we demonstrate the use of tags (evolvable labels that can be specified with imperfect matching) to identify memory positions in genetic programming (GP). Specifically, we conducted a series of experiments using simple linear-GP representations on five problems from the general program-synthesis benchmark suite [2]. We show that tag-indexed memory does not substantively affect problem solving success relative to more traditional, direct-indexed memory.

[1]  Lee Spector,et al.  Solving Uncompromising Problems With Lexicase Selection , 2015, IEEE Transactions on Evolutionary Computation.

[2]  Victor Ciesielski,et al.  Linear genetic programming , 2008, Genetic Programming and Evolvable Machines.

[3]  Hiroaki Kitano,et al.  Biological robustness , 2008, Nature Reviews Genetics.

[4]  Lee Spector,et al.  General Program Synthesis Benchmark Suite , 2015, GECCO.

[5]  Charles Ofria,et al.  What Else Is in an Evolved Name? Exploring Evolvable Specificity with SignalGP , 2018, GPTP.

[6]  Lee Spector,et al.  Tag-based modules in genetic programming , 2011, GECCO '11.