Abstract
Interpretability has drawn increasing attention in machine learning. Most works focus on post-hoc explanations rather than building a self-explaining model. So, we propose a Neural Partially Linear Additive Model (NPLAM), which automatically distinguishes insignificant, linear, and nonlinear features in neural networks. On the one hand, neural network construction fits data better than spline function under the same parameter amount; on the other hand, learnable gate design and sparsity regular-term maintain the ability of feature selection and structure discovery. We theoretically establish the generalization error bounds of the proposed method with Rademacher complexity. Experiments based on both simulations and real-world datasets verify its good performance and interpretability.
Similar content being viewed by others
References
Rudin C, Chen C, Chen Z, Huang H, Semenova L, Zhong C. Interpretable machine learning: fundamental principles and 10 grand challenges. Statistics Surveys, 2022, 16: 1–85
Du M, Liu N, Hu X. Techniques for interpretable machine learning. Communications of the ACM, 2019, 63(1): 68–77
Rudin C. Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead. Nature Machine Intelligence, 2019, 1(5): 206–215
Härdle W, Liang H, Gao J T. Partially Linear Models. Heidelberg: Physica, 2000
Xie Q, Liu J. Combined nonlinear effects of economic growth and urbanization on CO2 emissions in China: evidence from a panel data partially linear additive model. Energy, 2019, 186: 115868
Shim J H, Lee Y K. Generalized partially linear additive models for credit scoring. The Korean Journal of Applied Statistics, 2011, 24(4): 587–595
Kazemi M, Shahsavani D, Arashi M. Variable selection and structure identification for ultrahigh-dimensional partially linear additive models with application to cardiomyopathy microarray data. Statistics, Optimization & Information Computing, 2018, 6(3): 373–382
Zhang H H, Cheng G, Liu Y. Linear or nonlinear? Automatic structure discovery for partially linear models. Journal of the American Statistical Association, 2011, 106(495): 1099–1112
Du P, Cheng G, Liang H. Semiparametric regression models with additive nonparametric components and high dimensional parametric components. Computational Statistics & Data Analysis, 2012, 56(6): 2006–2017
Huang J, Wei F, Ma S. Semiparametric regression pursuit. Statistica Sinica, 2012, 22(4): 1403–1426
Lou Y, Bien J, Caruana R, Gehrke J. Sparse partially linear additive models. Journal of Computational and Graphical Statistics, 2016, 25(4): 1126–1140
Petersen A, Witten D. Data-adaptive additive modeling. Statistics in Medicine, 2019, 38(4): 583–600
Sadhanala V, Tibshirani R J. Additive models with trend filtering. The Annals of Statistics, 2019, 47(6): 3032–3068
Agarwal R, Melnick L, Frosst N, Zhang X, Lengerich B, Caruana R, Hinton G E. Neural additive models: Interpretable machine learning with neural nets. In: Proceedings of the 35th International Conference on Neural Information Processing Systems. 2021, 4699–4711
Nelder J A, Wedderburn R W M. Generalized linear models. Journal of the Royal Statistical Society. Series A (General), 1972, 135(3): 370–384
Hastie T, Tibshirani R. Generalized additive models. Statistical Science, 1986, 1(3): 297–310
Tibshirani R. Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society. Series B (Methodological), 1996, 58(1): 267–288
Ravikumar P, Lafferty J, Liu H, Wasserman L. Sparse additive models. Journal of the Royal Statistical Society. Series B (Statistical Methodology), 2009, 71(5): 1009–1030
Xu S Y, Bu Z Q, Chaudhari P, Barnett I J. Sparse neural additive model: Interpretable deep learning with feature selection via group sparsity. In: Proceedings of ICLR 2022 PAIR2Struct Workshop. 2022
Feng J, Simon N. Sparse-input neural networks for high-dimensional nonparametric regression and classification. 2017, arXiv preprint arXiv: 1711.07592v1
Lemhadri I, Ruan F, Abraham L, Tibshirani R. Lassonet: A neural network with feature sparsity. The Journal of Machine Learning Research, 2021, 22(1): 127
Wang X, Chen H, Yan J, Nho K, Risacher S L, Saykin A J, Shen L, Huang H, ADNI. Quantitative trait loci identification for brain endophenotypes via new additive model with random networks. Bioinformatics, 2018, 34(17): i866–i874
Nair V, Hinton G E. Rectified linear units improve restricted Boltzmann machines. In: Proceedings of the 27th International Conference on International Conference on Machine Learning. 2010, 807–814
Huber P J. Robust estimation of a location parameter. In: Kotz S, Johnson N L, eds. Breakthroughs in statistics: Methodology and Distribution. New York: Springer, 1992, 492–518
Lu Y Y, Fan Y, Lv J, Noble W S. DeepPINK: reproducible feature selection in deep neural networks. In: Proceedings of the 32nd International Conference on Neural Information Processing Systems. 2018, 8690–8700
Kingma D P, Ba J. Adam: A method for stochastic optimization. In: Proceedings of the 3rd International Conference on Learning Representations. 2015
Golowich N, Rakhlin A, Shamir O. Size-independent sample complexity of neural networks. In: Proceedings of Conference on Learning Theory. 2018, 297–299
McDiarmid C. On the method of bounded differences. In: Siemons J, ed. Surveys in Combinatorics. Cambridge: Cambridge University Press, 1989, 148–188
Chen H, Wang Y, Zheng F, Deng C, Huang H. Sparse modal additive model. IEEE Transactions on Neural Networks and Learning Systems, 2021, 32(6): 2373–2387
Wang X, Chen H, Cai W, Shen D, Huang H. Regularized modal regression with applications in cognitive impairment prediction. In: Proceedings of the 31st International Conference on Neural Information Processing Systems. 2017, 1447–1457
Cucker F, Zhou D X. Learning Theory: An Approximation Theory Viewpoint. Cambridge: Cambridge University Press, 2007
Wu Q, Ying Y, Zhou D X. Learning rates of least-square regularized regression. Foundations of Computational Mathematics, 2006, 6(2): 171–192
Krogh A. What are artificial neural networks? Nature Biotechnology, 2008, 26(2): 195–197
Ng A Y. Feature selection, L1 vs. L2 regularization, and rotational invariance. In: Proceedings of the 21st International Conference on Machine Learning. 2004, 78
Yang L, Lv S, Wang J. Model-free variable selection in reproducing kernel Hilbert space. The Journal of Machine Learning Research, 2016, 17(1): 2885–2908
Aygun R C, Yavuz A G. Network anomaly detection with stochastically improved autoencoder based models. In: Proceedings of the 4th IEEE International Conference on Cyber Security and Cloud Computing. 2017, 193–198
Chicco D, Warrens M J, Jurman G. The coefficient of determination R-squared is more informative than SMAPE, MAE, MAPE, MSE and RMSE in regression analysis evaluation. PeerJ Computer Science, 2021, 7: e623
Lin Y, Tu Y, Dou Z. An improved neural network pruning technology for automatic modulation classification in edge devices. IEEE Transactions on Vehicular Technology, 2020, 69(5): 5703–5706
Pace R K, Barry R. Sparse spatial autoregressions. Statistics & Probability Letters, 1997, 33(3): 291–297
Hamidieh K. A data-driven statistical model for predicting the critical temperature of a superconductor. Computational Materials Science, 2018, 154: 346–354
Zhang S, Guo B, Dong A, He J, Xu Z, Chen S X. Cautionary tales on air-quality improvement in Beijing. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 2017, 473(2205): 20170457
Harrison D Jr, Rubinfeld D L. Hedonic housing prices and the demand for clean air. Journal of Environmental Economics and Management, 1978, 5(1): 81–102
Buitinck L, Louppe G, Blondel M, Pedregosa F, Müeller A, Grisel O, Niculae V, Prettenhofer P, Gramfort A, Grobler J, Layton R, VanderPlas J, Joly A, Holt B, Varoquaux G. API design for machine learning software: experiences from the scikit-learn project. 2013, arXiv preprint arXiv: 1309.0238
Asuncion A, Newman D J. UCI machine learning repository. Irvine: Irvine University of California, 2017
Hazan E, Singh K. Boosting for online convex optimization. In: Proceedings of the 38th International Conference on Machine Learning. 2021, 4140–4149
Couellan N. Probabilistic robustness estimates for feed-forward neural networks. Neural Networks, 2021, 142: 138–147
Konstantinov A V, Utkin L V. Interpretable machine learning with an ensemble of gradient boosting machines. Knowledge-Based Systems, 2021, 222: 106993
Xing Y F, Xu Y H, Shi M H, Lian Y X. The impact of PM2.5 on the human respiratory system. Journal of Thoracic Disease, 2016, 8(1): E69–E74
Oune N, Bostanabad R. Latent map Gaussian processes for mixed variable metamodeling. Computer Methods in Applied Mechanics and Engineering, 2021, 387: 114128
Bekkar A, Hssina B, Douzi S, Douzi K. Air-pollution prediction in smart city, deep learning approach. Journal of Big Data, 2021, 8(1): 161
Author information
Authors and Affiliations
Corresponding author
Additional information
Liangxuan Zhu received his BS degree from the North China Institute of Science and Technology, China in 2019. He is currently pursuing the MS degree at the College of Informatics, Huazhong Agricultural University, China. His current research interests lie in machine learning, including deep learning, learning theory and interpretability.
Han Li received BS degree in Mathematics and Applied Mathematics from Faculty of Mathematics and Computer Science, Hubei University, China in 2007. She received her PhD degree in the School of Mathematics and Statistics at Beihang University, China. She worked as a project assistant professor in the Department of Mechanical Engineering, Kyushu University, Japan. She now works as an associate professor in the College of Informatics, Huazhong Agricultural University, China. Her research interests include neural networks, learning theory and pattern recognition.
Xuelin Zhang received his BE degree from the China Agricultural University, China in 2019. He is currently a PhD student with the College of Science, Huazhong Agricultural University, China. His current research interests include robust machine learning and statistical learning theory.
Lingjuan Wu got her PhD degree in Microelectronics and Solid State Electronics from Peking University, China in 2013. She visited the University of California, San Diego, USA as a research scholar from 2010 to 2012. She is currently an associate professor with the College of Informatics, Huazhong Agricultural University, China. Her research interests are in machine learning and hardware security, including learning theory and machine learning based side channel analysis and hardware Trojan detection.
Hong Chen received the BS and PhD degrees from Hubei University, China in 2003 and 2009, respectively. He worked as a postdoc researcher at University of Texas, USA during 2016–2017. He is currently a professor with the College of Informatics, Huazhong Agricultural University, China. His current research interests include machine learning, statistical learning theory and approximation theory.
Electronic Supplementary Material
Rights and permissions
About this article
Cite this article
Zhu, L., Li, H., Zhang, X. et al. Neural partially linear additive model. Front. Comput. Sci. 18, 186334 (2024). https://doi.org/10.1007/s11704-023-2662-3
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s11704-023-2662-3