دورية أكاديمية
أعرض في EDS

Optimization of Sparsity-Constrained Neural Networks as a Mixed Integer Linear Program: NN2MILP

التفاصيل البيبلوغرافية
العنوان: Optimization of Sparsity-Constrained Neural Networks as a Mixed Integer Linear Program: NN2MILP
المؤلفون: Rosenhahn, Bodo
المصدر: Journal of Optimization Theory and Applications 199 (2023) ; Journal of Optimization Theory and Applications
بيانات النشر: Springer Science + Business Media
سنة النشر: 2023
المجموعة: Institutional Repository of Leibniz Universität Hannover
مصطلحات موضوعية: Feature selection, Mixed integer linear programming, Neural networks, Resource optimization, Sparse networks, ddc:330, ddc:510, ddc:000
الوصف: The literature has shown how to optimize and analyze the parameters of different types of neural networks using mixed integer linear programs (MILP). Building on these developments, this work presents an approach to do so for a McCulloch/Pitts and Rosenblatt neurons. As the original formulation involves a step-function, it is not differentiable, but it is possible to optimize the parameters of neurons, and their concatenation as a shallow neural network, by using a mixed integer linear program. The main contribution of this paper is to additionally enforce sparsity constraints on the weights and activations as well as on the amount of used neurons. Several experiments demonstrate that such constraints effectively prevent overfitting in neural networks, and ensure resource optimized models.
نوع الوثيقة: article in journal/newspaper
اللغة: English
تدمد: 0022-3239
العلاقة: ESSN:1573-2878; http://dx.doi.org/10.15488/16139Test; https://www.repo.uni-hannover.de/handle/123456789/16266Test
DOI: 10.15488/16139
الإتاحة: https://doi.org/10.15488/16139Test
https://doi.org/10.1007/s10957-023-02317-xTest
https://www.repo.uni-hannover.de/handle/123456789/16266Test
حقوق: CC BY 4.0 Unported ; https://creativecommons.org/licenses/by/4.0Test ; frei zugänglich
رقم الانضمام: edsbas.1BE5171
قاعدة البيانات: BASE
الوصف
تدمد:00223239
DOI:10.15488/16139