What’s in an Evolved Name? The Evolution of Modularity via Tag-Based Reference

Spector, Lee; Harrington, Kyle; Martin, B.; Helmuth, Thomas

doi:10.1007/978-1-4614-1770-5_1

Cited by 18 publications

(11 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The lawnmower problem, and other versions of the CCP, have been extensively studied in the field of GP, but a review of GP techniques is beyond the scope of this paper. However, we highlight an observation that was revealed for a version of the CCP with obstacles [35]. In this study the meanings of evolved behavioral modules were investigated, and it was found that optimal solutions contained modules that encoded short action sequences which were then iterated to solve the CCP.…”

Section: Dirty Regionmentioning

confidence: 78%

Robot coverage control by evolved neuromodulation

Harrington

Awa

Cussat‐Blanc

et al. 2013

The 2013 International Joint Conference on Neural Networks (IJCNN)

View full text Add to dashboard Cite

Abstract-An important connection between evolution and learning was made over a century ago and is now termed as the Baldwin effect. Learning acts as a guide for an evolutionary search process. In this study reinforcement learning agents are trained to solve the robot coverage control problem. These agents are improved by evolving neuromodulatory gene regulatory networks (GRN) that influence the learning and memory of agents. Agents trained by these neuromodulatory GRNs can consistently generalize better than agents trained with fixed parameter settings. This work introduces evolutionary GRN models into the context of neuromodulation and illustrates some of the benefits that stem from neuromodulatory GRNs.

show abstract

Section: Dirty Regionmentioning

confidence: 78%

Robot coverage control by evolved neuromodulation

Harrington

Awa

Cussat‐Blanc

et al. 2013

The 2013 International Joint Conference on Neural Networks (IJCNN)

View full text Add to dashboard Cite

show abstract

“…Several involved the random selection of subtrees to create modules: Koza (1992a) proposed the use of a subtree encapsulation operator, which consists of randomly selecting a subtree from an individual to create a terminal primitive that encapsulated the subtree; Pollack (1992, 1994) proposed the Genetic Library Builder (GLiB) system, which employs mutation operators that randomly select subtrees to create modules (compress operator) that can be later expanded (expand operator); similar compress and expand operators to create and expand modules were more recently proposed by Walker and Miller (2008) in the context of Embedded Cartesian Genetic Programming (ECGP), with the extension of the use of module-altering operators (module point mutation, add-input, add-output, remove-input and remove-output operators); Spector et al (2011aSpector et al ( ,b, 2012 proposed the use of the concept of 'tags' to label fragments of code that can be later reused by referencing the same label-while this is similar to the use of a compress operator, it provides the flexibility of partial name matches (a label will match the closest matching tag).…”

Section: Modulatisation In Genetic Programmingmentioning

confidence: 99%

Sequential Symbolic Regression with Genetic Programming

Oliveira

Otero

Pappa

et al. 2015

Genetic and Evolutionary Computation

View full text Add to dashboard Cite

This chapter describes the Sequential Symbolic Regression (SSR) method, a new strategy for function approximation in symbolic regression. The SSR method is inspired by the sequential covering strategy from machine learning, but instead of sequentially reducing the size of the problem being solved, it sequentially transforms the original problem into potentially simpler problems. This transformation is performed according to the semantic distances between the desired and obtained outputs and a geometric semantic operator. The rationale behind SSR is that, after generating a suboptimal function f via symbolic regression, the output errors can be approximated by another function in a subsequent iteration. The method was tested in eight polynomial functions, and compared with canonical genetic programming (GP) and geometric semantic genetic programming (SGP). Results showed that SSR significantly outperforms SGP and presents no statistical difference to GP. More importantly, they show the potential of the proposed strategy: an effective way of applying geometric semantic operators to combine different (partial) solutions, avoiding the exponential growth problem arising from the use of these operators.

show abstract

“…The "closest match" semantics ensures that any reference to a tag will retrieve something as long as something has been tagged, and the number and targets of these references may grow gradually over evolutionary time. The results of previous studies have shown that this leads to problem-solving efforts for the lawnmower and "dirt-sensing, obstacle-avoiding robot" problems that scale well with problem size, indicating that tag-based modularity may be a powerful while nonetheless simple method for evolving modular programs more generally [24,21]. But because tree-based representations are currently more popular than Push among genetic programming researchers and practitioners, the extent to which such results can be obtained in tree-based genetic programming is clearly of interest.…”

Section: Tags In Pushmentioning

confidence: 99%

“…A recently-developed alternative technique for evolving programs that use modules, based on the use of "tags," has shown some promise insofar as it permits arbitrary architectures to evolve, even though the changes that must be made to the genetic programming system are relatively few and relatively simple [24,21]. The relevant notion of "tag" comes from the work of Holland, who used the term to denote initially arbitrary identifiers that come to have meaning over time in many different kinds of complex adaptive systems [5,6].…”

Section: Introductionmentioning

confidence: 99%

Tag-based modularity in tree-based genetic programming

Spector

Harrington

Helmuth

2012

Proceedings of the 14th Annual Conference on Genetic and Evolutionary Computation

Self Cite

View full text Add to dashboard Cite

Several techniques have been developed for allowing genetic programming systems to produce programs that make use of subroutines, macros, and other modular program structures. A recently proposed technique, based on the "tagging" and tag-based retrieval of blocks of code, has been shown to have novel and desirable features, but this was demonstrated only within the context of the PushGP genetic programming system. Following a suggestion in the GECCO-2011 publication on this technique we show here how tag-based modules can be incorporated into a more standard tree-based genetic programming system. We describe the technique in detail along with some possible extensions, outline arguments for its simplicity and potential power, and present results obtained using the technique on problems for which other modularization techniques have been shown to be useful. The results are mixed; substantial benefits are seen on the lawnmower problem but not on the Boolean even-4-parity problem. We discuss the observed results and directions for future research.

show abstract

What’s in an Evolved Name? The Evolution of Modularity via Tag-Based Reference

Cited by 18 publications

References 8 publications

Robot coverage control by evolved neuromodulation

Robot coverage control by evolved neuromodulation

Sequential Symbolic Regression with Genetic Programming

Tag-based modularity in tree-based genetic programming

Contact Info

Product

Resources

About