GINN-LP: A Growing Interpretable Neural Network for Discovering Multivariate Laurent Polynomial Equations

• DOI: 10.1609/aaai.v38i13.29396
• Publication Date: 2024-03-25T11:15:43Z
• Authors: Nisal Ranasinghe; Damith Senanayake; Sachith Seneviratne; Malin Premaratne; Saman Halgamuge
• Keywords: 0301 basic medicine; FOS: Computer and information sciences; Computer Science - Machine Learning; 03 medical and health sciences; Artificial Intelligence (cs.AI); Computer Science - Artificial Intelligence; Machine Learning (cs.LG)

Traditional machine learning is generally treated as a black-box optimization problem and does not typically produce interpretable functions that connect inputs outputs. However, the ability to discover such desirable. In this work, we propose GINN-LP, an neural network form coefficients of underlying equation dataset, when assumed take multivariate Laurent Polynomial. This facilitated by new type block, named “power-term approximator block”, consisting logarithmic exponential activation functions. GINN-LP end-to-end differentiable, making it possible use backpropagation for training. We growth strategy will enable finding suitable number terms in polynomial represents data, along with sparsity regularization promote discovery concise equations. To best our knowledge, first model can arbitrary without any prior information on order. Our approach evaluated subset data used SRBench, benchmark symbolic regression. show outperforms state-of-the-art regression methods datasets generated using 48 real-world equations polynomials. Next, ensemble method combines high-performing method, enabling us non-Laurent achieve results discovery, showing absolute improvement 7.1% over contender, applying 113 within SRBench known ground-truth