EXA-GP: Unifying Graph-Based Genetic Programming and Neuroevolution for Explainable Time Series Forecasting

Created by W.Langdon from gp-bibliography.bib Revision:1.8051

@InProceedings{murphy:2024:GECCOcomp,

• author = "Jared Murphy and Devroop Kar and Joshua Karns and Travis Desell",
• title = "{EXA-GP:} Unifying {Graph-Based} Genetic Programming and Neuroevolution for Explainable Time Series Forecasting",
• booktitle = "Proceedings of the 2024 Genetic and Evolutionary Computation Conference Companion",
• year = "2024",
• editor = "Ting Hu and Aniko Ekart",
• pages = "523--526",
• address = "Melbourne, Australia",
• series = "GECCO '24",
• month = "14-18 " # jul,
• organisation = "SIGEVO",
• publisher = "Association for Computing Machinery",
• publisher_address = "New York, NY, USA",
• keywords = "genetic algorithms, genetic programming, time series forecasting, graph-based genetic programming, neuroevolution, recurrent neural networks: Poster",
• isbn13 = "979-8-4007-0495-6",
• DOI = "doi:10.1145/3638530.3654349",
• size = "4 pages",
• abstract = "This work introduces Evolutionary eXploration of Augmenting Genetic Programs (EXA-GP), which converts the EXAMM neuroevolution algorithm into a graph-based genetic programming (GGP) algorithm by swapping out its library of neurons and recurrent memory cells for genetic programming (GP) operations. This enables EXA-GP to use the full suite of optimizations provided by EXAMM, such as distributed and multi-threaded execution, island based populations with re-population events, as well as Lamarckian weight inheritance and backpropagation for optimization of constant values. Results show that EXA-GP's evolved genetic programs achieve the same levels of accuracy as recurrent neural networks (RNNs) evolved by EXAMM, while at the same time being more explainable. EXA-GP's genetic programs are also computationally less complex than EXAMM's RNNs, suggesting that GP operations may actually be more effective than gated recurrent memory cells for time series forecasting. We also show that EXAMM and EXA-GP outperform state-of-the-art GP and recurrent CGPANN algorithms which struggle on these benchmarks.",
• notes = "GECCO-2024 GP A Recombination of the 33rd International Conference on Genetic Algorithms (ICGA) and the 29th Annual Genetic Programming Conference (GP)",

}

Genetic Programming entries for Jared Murphy Devroop Kar Joshua Karns Travis Desell

Citations