Exploring Hidden Semantics in Neural Networks with Symbolic Regression

Created by W.Langdon from gp-bibliography.bib Revision:1.7630

@InProceedings{luo:2022:GECCO,

• author = "Yuanzhen Luo and Qiang Lu and Xilei Hu and Jake Luo and Zhiguang Wang",
• title = "Exploring Hidden Semantics in Neural Networks with Symbolic Regression",
• booktitle = "Proceedings of the 2022 Genetic and Evolutionary Computation Conference",
• year = "2022",
• editor = "Alma Rahat and Jonathan Fieldsend and Markus Wagner and Sara Tari and Nelishia Pillay and Irene Moser and Aldeida Aleti and Ales Zamuda and Ahmed Kheiri and Erik Hemberg and Christopher Cleghorn and Chao-li Sun and Georgios Yannakakis and Nicolas Bredeche and Gabriela Ochoa and Bilel Derbel and Gisele L. Pappa and Sebastian Risi and Laetitia Jourdan and Hiroyuki Sato and Petr Posik and Ofer Shir and Renato Tinos and John Woodward and Malcolm Heywood and Elizabeth Wanner and Leonardo Trujillo and Domagoj Jakobovic and Risto Miikkulainen and Bing Xue and Aneta Neumann and Richard Allmendinger and Inmaculada Medina-Bulo and Slim Bechikh and Andrew M. Sutton and Pietro Simone Oliveto",
• pages = "982--990",
• address = "Boston, USA",
• series = "GECCO '22",
• month = "9-13 " # jul,
• organisation = "SIGEVO",
• publisher = "Association for Computing Machinery",
• publisher_address = "New York, NY, USA",
• keywords = "genetic algorithms, genetic programming, Cartesian Genetic Programming, neural network, ANN, symbolic regression, Lime, Maple, SRNet, USDB",
• isbn13 = "978-1-4503-9237-2",
• URL = "https://arxiv.org/abs/2204.10529",
• DOI = "doi:10.1145/3512290.3528758",
• size = "9 pages",
• abstract = "Many recent studies focus on developing mechanisms to explain the black-box behaviors of neural networks (NNs). However, little work has been done to extract the potential hidden semantics (mathematical representation) of a neural network. A succinct and explicit mathematical representation of a ANN model could improve the understanding and interpretation of its behaviors. To address this need, we propose a novel symbolic regression method for neural works (called SRNet) to discover the mathematical expressions of a NN. SRNet creates a Cartesian genetic programming (NNCGP) to represent the hidden semantics of a single layer in a NN. It then leverages a multi-chromosome NNCGP to represent hidden semantics of all layers of the NN. The method uses a (1+λ) evolutionary strategy (called MNNCGP-ES) to extract the final mathematical expressions of all layers in the NN. Experiments on 12 symbolic regression benchmarks and 5 classification benchmarks show that SRNet not only can reveal the complex relationships between each layer of a NN but also can extract the mathematical representation of the whole NN. Compared with LIME and MAPLE, SRNet has higher interpolation accuracy and trends to approximate the real model on the practical dataset",
• notes = "GECCO-2022 A Recombination of the 31st International Conference on Genetic Algorithms (ICGA) and the 27th Annual Genetic Programming Conference (GP)",

}

Genetic Programming entries for Yuanzhen Luo Qiang Lu Xilei Hu Jake Luo Zhiguang Wang

Citations