Genetic programming benchmarks

McDermott, James; Kronberger, Gabriel; Orzechowski, Patryk; Vanneschi, Leonardo; Manzoni, Luca; Kalkreuth, Roman; Castelli, Mauro

doi:10.1145/3578482.3578483

Cited by 7 publications

(4 citation statements)

References 23 publications

(25 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The recently proposed SRBench benchmark suite [3,4], presents the most comprehensive, systematic and reproducible evaluation process for SR algorithms. While SR is the most widely studied problem in GP, assessing and comparing SR methods was mostly done in an ad hoc manner before SRBench [47,48]. Moreover, it provides open-source and easy-to-use implementations of the state-of-the-art methods, considering both GP and non-GP methods.…”

Section: Related Workmentioning

confidence: 99%

M5GP: Parallel Multidimensional Genetic Programming with Multidimensional Populations for Symbolic Regression

Cárdenas Florido,

Trujillo,

Hernandez

et al. 2024

MCA

View full text Add to dashboard Cite

Machine learning and artificial intelligence are growing in popularity thanks to their ability to produce models that exhibit unprecedented performance in domains that include computer vision, natural language processing and code generation. However, such models tend to be very large and complex and impossible to understand using traditional analysis or human scrutiny. Conversely, Symbolic Regression methods attempt to produce models that are relatively small and (potentially) human-readable. In this domain, Genetic Programming (GP) has proven to be a powerful search strategy that achieves state-of-the-art performance. This paper presents a new GP-based feature transformation method called M5GP, which is hybridized with multiple linear regression to produce linear models, implemented to exploit parallel processing on graphical processing units for efficient computation. M5GP is the most recent variant from a family of feature transformation methods (M2GP, M3GP and M4GP) that have proven to be powerful tools for both classification and regression tasks applied to tabular data. The proposed method was evaluated on SRBench v2.0, the current standard benchmarking suite for Symbolic Regression. Results show that M5GP achieves performance that is competitive with the state-of-the-art, achieving a top-three rank on the most difficult subset of black-box problems. Moreover, it achieves the lowest computation time when compared to other GP-based methods that have similar accuracy scores.

show abstract

Section: Related Workmentioning

confidence: 99%

M5GP: Parallel Multidimensional Genetic Programming with Multidimensional Populations for Symbolic Regression

Cárdenas Florido,

Trujillo,

Hernandez

et al. 2024

MCA

View full text Add to dashboard Cite

show abstract

“…However, according to many authors who have used this dataset, these characteristics are interesting and should be integrated in a reasonable benchmark suite, because they allow us to test the ability of our algorithms to deal with the difficulties and ambiguities that are typical of real-world data. It is not our objective to discuss what characteristics a good benchmark suite should possess (the interested reader is referred to [45][46][47][48] for such a discussion). We simply observe that the Bioavailability dataset, as well as the PPB and LD50 datasets, have x, z = rnd(−1, 1), y = rnd(1, 2)…”

Section: Test Problemsmentioning

confidence: 99%

Geometric semantic GP with linear scaling: Darwinian versus Lamarckian evolution

Nadizar,

Sakallioglu,

Garrow

et al. 2024

Genet Program Evolvable Mach

Self Cite

View full text Add to dashboard Cite

Geometric Semantic Genetic Programming (GSGP) has shown notable success in symbolic regression with the introduction of Linear Scaling (LS). This achievement stems from the synergy of the geometric semantic genetic operators of GSGP with the scaling of the individuals for computing their fitness, which favours programs with a promising behaviour. However, the initial combination of GSGP and LS (GSGP-LS) underutilised the potential of LS, scaling individuals only for fitness evaluation, neglecting to incorporate improvements into their genetic material. In this paper we propose an advancement, GSGP with Lamarckian LS (GSGP-LLS), wherein we update the individuals in the population with their scaling coefficients in a Lamarckian fashion, i.e., by inheritance of acquired traits. We assess GSGP-LS and GSGP-LLS against standard GSGP for the task of symbolic regression on five hand-tailored benchmarks and six real-life problems. On the former ones, GSGP-LS and GSGP-LLS both consistently improve GSGP, though with no clear global superiority between them. On the real-world problems, instead, GSGP-LLS steadily outperforms GSGP-LS, achieving faster convergence and superior final performance. Notably, even in cases where LS induces overfitting on challenging problems, GSGP-LLS surpasses GSGP-LS, due to its slower and more localised optimisation steps.

show abstract

“…Moreover, the authors state that for the future of benchmarking in LS, it might be useful to further increase the diversity of benchmarks by exploring new Boolean function problems and curating these problems into a new benchmark suite. Therefore, in this work, we follow up the suggestion of McDermott et al [19] and consider benchmarking in LS from a general perspective. We reflect on the requirements for a general benchmark suite for LS, and bundle together a set of Boolean functions from the major categories commonly used in previous work on GP.…”

Section: Introductionmentioning

confidence: 98%

“…At last year's GECCO conference, the work of McDermott et al [20] was awarded the SIGEVO Impact award, which triggered reflection on the developments of the last decades. Very recently, a followup article on the state and of development of Benchmarking has been published by McDermott et al [19]. Besides reviewing wellestablished GP benchmark suites which have been proposed over the last years, the missing of a Boolean function benchmark suite for logic synthesis (LS) has been identified as one of the major gaps.…”

Section: Introductionmentioning

confidence: 99%

Towards a General Boolean Function Benchmark Suite

Kalkreuth¹,

Vašíček

Husa

et al. 2023

Proceedings of the Companion Conference on Genetic and Evolutionary Computation

Self Cite

View full text Add to dashboard Cite

over a decade ago, the first comprehensive review on the state of benchmarking in Genetic Programming (GP) analyzed the mismatch between the problems that are used to test the performance of GP systems and real-world problems. Since then, several benchmark suites in major GP problem domains have been proposed over time, which were able to fill some of the major gaps. In the framework of the first review about the state of benchmarking in GP, logic synthesis was classified as one of the major GP problem domains. However, a diverse and accessible benchmark suite for logic synthesis is still missing in the field of GP. In this work, we take a first step towards a benchmark suite for logic synthesis that covers different types of Boolean functions that are commonly used for the evaluation of GP systems. We also present baseline results that have been obtained by former work and in our evaluation experiments by using Cartesian Genetic Programming.

show abstract

Genetic programming benchmarks

Cited by 7 publications

References 23 publications

M5GP: Parallel Multidimensional Genetic Programming with Multidimensional Populations for Symbolic Regression

M5GP: Parallel Multidimensional Genetic Programming with Multidimensional Populations for Symbolic Regression

Geometric semantic GP with linear scaling: Darwinian versus Lamarckian evolution

Towards a General Boolean Function Benchmark Suite

Contact Info

Product

Resources

About