Measuring bloat, overfitting and functional complexity in genetic programming

Vanneschi, Leonardo; Castelli, Mauro; Silva, Sara

doi:10.1145/1830483.1830643

Cited by 108 publications

(122 citation statements)

References 21 publications

Supporting

Mentioning

118

Contrasting

Order By: Relevance

“…The goal of the experimental work is to test the NS-GP classifier on two-class classification problems, evaluating its performance based on classification error and the mean size of the evolved population, a good indicator of the effect bloating is having on a GP run [37]. The proposed algorithm will hereafter be denoted by NS-SRS.…”

Section: Experiments and Resultsmentioning

confidence: 99%

Preliminary Study of Bloat in Genetic Programming with Behavior-Based Search

Trujillo

Naredo

Martínez

2013

EVOLVE - A Bridge Between Probability, Set Oriented Numerics, and Evolutionary Computation IV

View full text Add to dashboard Cite

Abstract. Bloat is one of the most interesting theoretical problems in genetic programming (GP), and one of the most important pragmatic limitations in the development of real-world GP solutions. Over the years, many theories regarding the causes of bloat have been proposed and a variety of bloat control methods have been developed. It seems that one of the underlying causes of bloat is the search for fitness; as the fitness-causes-bloat theory states, selective bias towards fitness seems to unavoidably lead the search towards programs with a large size. Intuitively, however, abandoning fitness does not appear to be an option. This paper, studies a GP system that does not require an explicit fitness function, instead it relies on behavior-based search, where programs are described by the behavior they exhibit and selective pressure is biased towards unique behaviors using the novelty search algorithm. Initial results are encouraging, the average program size of the evolving population does not increase with novelty search; i.e., bloat is avoided by focusing on novelty instead of quality.

show abstract

Section: Experiments and Resultsmentioning

confidence: 99%

Preliminary Study of Bloat in Genetic Programming with Behavior-Based Search

Trujillo

Naredo

Martínez

2013

EVOLVE - A Bridge Between Probability, Set Oriented Numerics, and Evolutionary Computation IV

View full text Add to dashboard Cite

show abstract

“…Therefore, bloat was assumed to be a good indicator of program overfitting. However, recent experimental work suggests that this assumption is not reliable [14]. In particular, [14] showed that a causal link between bloat and overfitting did not exist on three test cases.…”

Section: Introductionmentioning

confidence: 95%

“…However, recent experimental work suggests that this assumption is not reliable [14]. In particular, [14] showed that a causal link between bloat and overfitting did not exist on three test cases. From this it follows that bloated programs should not be a priori regarded as complex.…”

Section: Introductionmentioning

confidence: 95%

“…Moreover, it is also assumed that bloat is related to overfitting, one of the most important problems in machine learning. It is often stated that simpler solutions will be more robust and will generalize better than complex ones, with the latter being more likely to overfit the training data [6,14]. The GP community has tacitly assumed that simple solutions can be equated with small program trees, and that very large programs correspond to complex solutions [8].…”

Section: Introductionmentioning

confidence: 99%

“…And second, can program complexity be used as an indicator of program overfitting? Here, we study the measure of functional complexity proposed in [14], which we call Slope-based Functional Complexity (SFC). SFC in inspired by the concept of curvature, and represents the first measure of complexity that is explicitly intended to be an indicator of program overfitting.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

An Empirical Study of Functional Complexity as an Indicator of Overfitting in Genetic Programming

Trujillo

Silva

Legrand

et al. 2011

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

Abstract.Recently, it has been stated that the complexity of a solution is a good indicator of the amount of overfitting it incurs. However, measuring the complexity of a program, in Genetic Programming, is not a trivial task. In this paper, we study the functional complexity and how it relates with overfitting on symbolic regression problems. We consider two measures of complexity, Slope-based Functional Complexity, inspired by the concept of curvature, and Regularity-based Functional Complexity based on the concept of Hölderian regularity. In general, both complexity measures appear to be poor indicators of program overfitting. However, results suggest that Regularity-based Functional Complexity could provide a good indication of overfitting in extreme cases.

show abstract

Swarm‐based metaheuristics in automatic programming: a survey

Olmo

Romero

Ventura

2014

WIREs Data Min & Knowl

View full text Add to dashboard Cite

On the one hand, swarm intelligence (SI) is an emerging field of artificial intelligence that takes inspiration in the collective and social behavior of different groups of simple agents. On the other hand, the automatic evolution of programs is an active research area that has attracted a lot of interest and has been mostly promoted by the genetic programming paradigm. The main objective is to find computer programs from a high‐level problem statement of what needs to be done, without needing to know the structure of the solution beforehand. This paper looks at the intersection between SI and automatic programming, providing a survey on the state‐of‐the‐art of the automatic programming algorithms that use an SI metaheuristic as the search technique. The expression of swarm programming (SP) has been coined to cover swarm‐based automatic programming proposals, since they have been published to date in a disorganized manner. Open issues for future research are listed. Although it is a very recent area, we hope that this work will stimulate the interest of the research community in the development of new SP metaheuristics, algorithms, and applications. WIREs Data Mining Knowl Discov 2014, 4:445–469. doi: 10.1002/widm.1138 This article is categorized under: Algorithmic Development > Association Rules Technologies > Classification Technologies > Computational Intelligence

show abstract

Measuring bloat, overfitting and functional complexity in genetic programming

Cited by 108 publications

References 21 publications

Preliminary Study of Bloat in Genetic Programming with Behavior-Based Search

Preliminary Study of Bloat in Genetic Programming with Behavior-Based Search

An Empirical Study of Functional Complexity as an Indicator of Overfitting in Genetic Programming

Swarm‐based metaheuristics in automatic programming: a survey

Contact Info

Product

Resources

About