[{"title":"Multi-Objective Optimization for Sparse Deep Multi-Task Learning","date_created":"2023-08-24T07:44:36Z","publisher":"IEEE","year":"2024","language":[{"iso":"eng"}],"abstract":[{"text":"Different conflicting optimization criteria arise naturally in various Deep\r\nLearning scenarios. These can address different main tasks (i.e., in the\r\nsetting of Multi-Task Learning), but also main and secondary tasks such as loss\r\nminimization versus sparsity. The usual approach is a simple weighting of the\r\ncriteria, which formally only works in the convex setting. In this paper, we\r\npresent a Multi-Objective Optimization algorithm using a modified Weighted\r\nChebyshev scalarization for training Deep Neural Networks (DNNs) with respect\r\nto several tasks. By employing this scalarization technique, the algorithm can\r\nidentify all optimal solutions of the original problem while reducing its\r\ncomplexity to a sequence of single-objective problems. The simplified problems\r\nare then solved using an Augmented Lagrangian method, enabling the use of\r\npopular optimization techniques such as Adam and Stochastic Gradient Descent,\r\nwhile efficaciously handling constraints. Our work aims to address the\r\n(economical and also ecological) sustainability issue of DNN models, with a\r\nparticular focus on Deep Multi-Task models, which are typically designed with a\r\nvery large number of weights to perform equally well on multiple tasks. Through\r\nexperiments conducted on two Machine Learning datasets, we demonstrate the\r\npossibility of adaptively sparsifying the model during training without\r\nsignificantly impacting its performance, if we are willing to apply\r\ntask-specific adaptations to the network weights. Code is available at\r\nhttps://github.com/salomonhotegni/MDMTN.","lang":"eng"}],"publication":"2024 International Joint Conference on Neural Networks (IJCNN)","main_file_link":[{"open_access":"1","url":"https://ieeexplore.ieee.org/document/10650994"}],"doi":"10.1109/IJCNN60899.2024.10650994","conference":{"end_date":"2024-07-05","location":"Yokohama, Japan","name":"2024 International Joint Conference on Neural Networks (IJCNN)","start_date":"2024-06-30"},"author":[{"last_name":"Hotegni","full_name":"Hotegni, Sedjro Salomon","id":"97995","first_name":"Sedjro Salomon"},{"id":"51701","full_name":"Berkemeier, Manuel Bastian","last_name":"Berkemeier","first_name":"Manuel Bastian"},{"id":"47427","full_name":"Peitz, Sebastian","orcid":"0000-0002-3389-793X","last_name":"Peitz","first_name":"Sebastian"}],"date_updated":"2024-09-27T10:24:22Z","oa":"1","citation":{"bibtex":"@inproceedings{Hotegni_Berkemeier_Peitz_2024, place={Yokohama, Japan}, title={Multi-Objective Optimization for Sparse Deep Multi-Task Learning}, DOI={<a href=\"https://doi.org/10.1109/IJCNN60899.2024.10650994\">10.1109/IJCNN60899.2024.10650994</a>}, booktitle={2024 International Joint Conference on Neural Networks (IJCNN)}, publisher={IEEE}, author={Hotegni, Sedjro Salomon and Berkemeier, Manuel Bastian and Peitz, Sebastian}, year={2024}, pages={9} }","mla":"Hotegni, Sedjro Salomon, et al. “Multi-Objective Optimization for Sparse Deep Multi-Task Learning.” <i>2024 International Joint Conference on Neural Networks (IJCNN)</i>, IEEE, 2024, p. 9, doi:<a href=\"https://doi.org/10.1109/IJCNN60899.2024.10650994\">10.1109/IJCNN60899.2024.10650994</a>.","short":"S.S. Hotegni, M.B. Berkemeier, S. Peitz, in: 2024 International Joint Conference on Neural Networks (IJCNN), IEEE, Yokohama, Japan, 2024, p. 9.","apa":"Hotegni, S. S., Berkemeier, M. B., &#38; Peitz, S. (2024). Multi-Objective Optimization for Sparse Deep Multi-Task Learning. <i>2024 International Joint Conference on Neural Networks (IJCNN)</i>, 9. <a href=\"https://doi.org/10.1109/IJCNN60899.2024.10650994\">https://doi.org/10.1109/IJCNN60899.2024.10650994</a>","ama":"Hotegni SS, Berkemeier MB, Peitz S. Multi-Objective Optimization for Sparse Deep Multi-Task Learning. In: <i>2024 International Joint Conference on Neural Networks (IJCNN)</i>. IEEE; 2024:9. doi:<a href=\"https://doi.org/10.1109/IJCNN60899.2024.10650994\">10.1109/IJCNN60899.2024.10650994</a>","ieee":"S. S. Hotegni, M. B. Berkemeier, and S. Peitz, “Multi-Objective Optimization for Sparse Deep Multi-Task Learning,” in <i>2024 International Joint Conference on Neural Networks (IJCNN)</i>, Yokohama, Japan, 2024, p. 9, doi: <a href=\"https://doi.org/10.1109/IJCNN60899.2024.10650994\">10.1109/IJCNN60899.2024.10650994</a>.","chicago":"Hotegni, Sedjro Salomon, Manuel Bastian Berkemeier, and Sebastian Peitz. “Multi-Objective Optimization for Sparse Deep Multi-Task Learning.” In <i>2024 International Joint Conference on Neural Networks (IJCNN)</i>, 9. Yokohama, Japan: IEEE, 2024. <a href=\"https://doi.org/10.1109/IJCNN60899.2024.10650994\">https://doi.org/10.1109/IJCNN60899.2024.10650994</a>."},"page":"9","place":"Yokohama, Japan","publication_status":"published","has_accepted_license":"1","publication_identifier":{"eisbn":["979-8-3503-5931-2"],"eissn":[" 2161-4407"]},"user_id":"97995","department":[{"_id":"655"}],"_id":"46649","status":"public","type":"conference"},{"department":[{"_id":"101"},{"_id":"655"}],"user_id":"47427","external_id":{"arxiv":["2208.12094"]},"_id":"33150","language":[{"iso":"eng"}],"publication":"arXiv:2208.12094","type":"preprint","status":"public","abstract":[{"lang":"eng","text":"In this article, we build on previous work to present an optimization algorithm for nonlinearly constrained multi-objective optimization problems. The algorithm combines a surrogate-assisted derivative-free trust-region approach with the filter method known from single-objective optimization. Instead of the true objective and constraint functions, so-called fully linear models are employed and we show how to deal with the gradient inexactness in the composite step setting, adapted from single-objective optimization as well. Under standard assumptions, we prove convergence of a subset of iterates to a quasi-stationary point and if constraint qualifications hold, then the limit point is also a KKT-point of the multi-objective problem."}],"author":[{"last_name":"Berkemeier","full_name":"Berkemeier, Manuel Bastian","id":"51701","first_name":"Manuel Bastian"},{"last_name":"Peitz","orcid":"0000-0002-3389-793X","full_name":"Peitz, Sebastian","id":"47427","first_name":"Sebastian"}],"date_created":"2022-08-26T06:08:06Z","oa":"1","date_updated":"2022-08-26T06:12:10Z","main_file_link":[{"url":"https://arxiv.org/pdf/2208.12094","open_access":"1"}],"title":"Multi-Objective Trust-Region Filter Method for Nonlinear Constraints using Inexact Gradients","citation":{"short":"M.B. Berkemeier, S. Peitz, ArXiv:2208.12094 (2022).","mla":"Berkemeier, Manuel Bastian, and Sebastian Peitz. “Multi-Objective Trust-Region Filter Method for Nonlinear Constraints Using Inexact Gradients.” <i>ArXiv:2208.12094</i>, 2022.","bibtex":"@article{Berkemeier_Peitz_2022, title={Multi-Objective Trust-Region Filter Method for Nonlinear Constraints using Inexact Gradients}, journal={arXiv:2208.12094}, author={Berkemeier, Manuel Bastian and Peitz, Sebastian}, year={2022} }","apa":"Berkemeier, M. B., &#38; Peitz, S. (2022). Multi-Objective Trust-Region Filter Method for Nonlinear Constraints using Inexact Gradients. In <i>arXiv:2208.12094</i>.","ama":"Berkemeier MB, Peitz S. Multi-Objective Trust-Region Filter Method for Nonlinear Constraints using Inexact Gradients. <i>arXiv:220812094</i>. Published online 2022.","ieee":"M. B. Berkemeier and S. Peitz, “Multi-Objective Trust-Region Filter Method for Nonlinear Constraints using Inexact Gradients,” <i>arXiv:2208.12094</i>. 2022.","chicago":"Berkemeier, Manuel Bastian, and Sebastian Peitz. “Multi-Objective Trust-Region Filter Method for Nonlinear Constraints Using Inexact Gradients.” <i>ArXiv:2208.12094</i>, 2022."},"year":"2022"},{"date_created":"2021-03-01T10:46:48Z","title":"Derivative-Free Multiobjective Trust Region Descent Method Using Radial  Basis Function Surrogate Models","issue":"2","year":"2021","language":[{"iso":"eng"}],"publication":"Mathematical and Computational Applications","abstract":[{"text":"We present a flexible trust region descend algorithm for unconstrained and\r\nconvexly constrained multiobjective optimization problems. It is targeted at\r\nheterogeneous and expensive problems, i.e., problems that have at least one\r\nobjective function that is computationally expensive. The method is\r\nderivative-free in the sense that neither need derivative information be\r\navailable for the expensive objectives nor are gradients approximated using\r\nrepeated function evaluations as is the case in finite-difference methods.\r\nInstead, a multiobjective trust region approach is used that works similarly to\r\nits well-known scalar pendants. Local surrogate models constructed from\r\nevaluation data of the true objective functions are employed to compute\r\npossible descent directions. In contrast to existing multiobjective trust\r\nregion algorithms, these surrogates are not polynomial but carefully\r\nconstructed radial basis function networks. This has the important advantage\r\nthat the number of data points scales linearly with the parameter space\r\ndimension. The local models qualify as fully linear and the corresponding\r\ngeneral scalar framework is adapted for problems with multiple objectives.\r\nConvergence to Pareto critical points is proven and numerical examples\r\nillustrate our findings.","lang":"eng"}],"oa":"1","date_updated":"2022-01-06T06:54:55Z","author":[{"first_name":"Manuel Bastian","full_name":"Berkemeier, Manuel Bastian","id":"51701","last_name":"Berkemeier"},{"first_name":"Sebastian","last_name":"Peitz","orcid":"0000-0002-3389-793X","full_name":"Peitz, Sebastian","id":"47427"}],"volume":26,"main_file_link":[{"url":"https://www.mdpi.com/2297-8747/26/2/31/pdf","open_access":"1"}],"doi":"10.3390/mca26020031","publication_status":"published","publication_identifier":{"eissn":["2297-8747"]},"citation":{"ieee":"M. B. Berkemeier and S. Peitz, “Derivative-Free Multiobjective Trust Region Descent Method Using Radial  Basis Function Surrogate Models,” <i>Mathematical and Computational Applications</i>, vol. 26, no. 2, 2021.","chicago":"Berkemeier, Manuel Bastian, and Sebastian Peitz. “Derivative-Free Multiobjective Trust Region Descent Method Using Radial  Basis Function Surrogate Models.” <i>Mathematical and Computational Applications</i> 26, no. 2 (2021). <a href=\"https://doi.org/10.3390/mca26020031\">https://doi.org/10.3390/mca26020031</a>.","ama":"Berkemeier MB, Peitz S. Derivative-Free Multiobjective Trust Region Descent Method Using Radial  Basis Function Surrogate Models. <i>Mathematical and Computational Applications</i>. 2021;26(2). doi:<a href=\"https://doi.org/10.3390/mca26020031\">10.3390/mca26020031</a>","apa":"Berkemeier, M. B., &#38; Peitz, S. (2021). Derivative-Free Multiobjective Trust Region Descent Method Using Radial  Basis Function Surrogate Models. <i>Mathematical and Computational Applications</i>, <i>26</i>(2). <a href=\"https://doi.org/10.3390/mca26020031\">https://doi.org/10.3390/mca26020031</a>","bibtex":"@article{Berkemeier_Peitz_2021, title={Derivative-Free Multiobjective Trust Region Descent Method Using Radial  Basis Function Surrogate Models}, volume={26}, DOI={<a href=\"https://doi.org/10.3390/mca26020031\">10.3390/mca26020031</a>}, number={231}, journal={Mathematical and Computational Applications}, author={Berkemeier, Manuel Bastian and Peitz, Sebastian}, year={2021} }","mla":"Berkemeier, Manuel Bastian, and Sebastian Peitz. “Derivative-Free Multiobjective Trust Region Descent Method Using Radial  Basis Function Surrogate Models.” <i>Mathematical and Computational Applications</i>, vol. 26, no. 2, 31, 2021, doi:<a href=\"https://doi.org/10.3390/mca26020031\">10.3390/mca26020031</a>.","short":"M.B. Berkemeier, S. Peitz, Mathematical and Computational Applications 26 (2021)."},"intvolume":"        26","_id":"21337","user_id":"47427","department":[{"_id":"101"},{"_id":"655"}],"article_number":"31","type":"journal_article","status":"public"}]