{"type":"ScholarlyArticle","title":"Robust Losses for Decision-Focused Learning","authors":[{"name":"Noah Schutte","slug":"noah-schutte","given":"Noah","family":"Schutte","affiliation":"Delft University of Technology"},{"name":"Krzysztof Postek","slug":"krzysztof-postek","given":"Krzysztof","family":"Postek","affiliation":"Independent Researcher"},{"name":"Neil Yorke-Smith","slug":"neil-yorke-smith","given":"Neil","family":"Yorke-Smith","affiliation":"Delft University of Technology"}],"abstract":"Optimization models used to make discrete decisions often contain uncertain parameters that are context-dependent and estimated through prediction. To account for the quality of the decision made based on the prediction, decision-focused learning (end-to-end predict-then-optimize) aims at training the predictive model to minimize regret, i.e., the loss incurred by making a suboptimal decision. Despite the challenge of the gradient of this loss w.r.t. the predictive model parameters being zero almost everywhere for optimization problems with a linear objective, effective gradient-based learning approaches have been proposed to minimize the expected loss, using the empirical loss as a surrogate. However, empirical regret can be an ineffective surrogate because empirical optimal decisions can vary substantially from expected optimal decisions. To understand the impact of this deficiency, we evaluate the effect of aleatoric and epistemic uncertainty on the accuracy of empirical regret as a surrogate. Next, we propose three novel loss functions that approximate expected regret more robustly. Experimental results show that training two state-of-the-art decision-focused learning approaches using robust regret losses improves test-sample empirical regret in general while keeping computational time equivalent relative to the number of training epochs.","doi":"10.24963/ijcai.2024/538","url":"https://ijcai.org/proceedings/2024/538","pdf":"https://ijcai.org/proceedings/2024/0538.pdf","markdown":"https://ijcai.org/proceedings/2024/538.md","bibtex":"https://ijcai.org/proceedings/2024/bibtex/538","volume":{"slug":"2024","title":"Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence","short_title":"IJCAI 2024","url":"https://ijcai.org/proceedings/2024/","year":2024,"location":"Jeju, Korea","dates":"3-9 August 2024","editors":"Kate Larson","isbn":"978-1-956792-04-1","publisher":"International Joint Conferences on Artificial Intelligence Organization","doi_prefix":"10.24963/ijcai.2024"},"section":"Machine Learning","track":"Main Track","number":538,"pages":{"first":4868,"last":4875},"keywords":["Machine Learning: ML: Robustness","Machine Learning: ML: Regression","Machine Learning: ML: Optimization","Constraint Satisfaction and Optimization: CSO: Constraint optimization problems"],"videos":[],"references":[{"position":1,"raw":"[Amos and Kolter, 2017] Brandon Amos and J. Zico Kolter. OptNet: Differentiable optimization as a layer in neural networks. International Conference on Machine Learning, PMLR, 70:136-145, 2017.","title":"OptNet: Differentiable optimization as a layer in neural networks","authors":"Brandon Amos, J Zico Kolter","year":"2017","venue":"International Conference on Machine Learning","doi":null,"url":null,"internal_url":null},{"position":2,"raw":"[Bengio, 1997] Yoshua Bengio. Using a financial training criterion rather than a prediction criterion. International Journal of Neural Systems, 8:433-443, 1997.","title":"Using a financial training criterion rather than a prediction criterion","authors":"Yoshua Bengio","year":"1997","venue":"International Journal of Neural Systems","doi":null,"url":null,"internal_url":null},{"position":3,"raw":"[Berthet et al., 2020] Quentin Berthet, Mathieu Blondel, Olivier Teboul, Marco Cuturi, Jean-Philippe Vert, and Fran- cis Bach. Learning with differentiable perturbed optimiz- ers. Advances in Neural Information Processing Systems, 33:9508-9519, 2020.","title":"Learning with differentiable perturbed optimizers","authors":"Quentin Berthet, Mathieu Blondel, Olivier Teboul, Marco Cuturi, Jean-Philippe Vert, Francis Bach","year":"2020","venue":"Advances in Neural Information Processing Systems","doi":null,"url":null,"internal_url":null},{"position":4,"raw":"[Bertsimas and den Hertog, 2022] Dimitris Bertsimas and Dick den Hertog. Robust and adaptive optimization. Dy- namic Ideas LLC, 2022.","title":"Dimitris Bertsimas and Dick den Hertog. Robust and adaptive optimization","authors":"Den Bertsimas, Hertog","year":"2022","venue":null,"doi":null,"url":null,"internal_url":null},{"position":5,"raw":"[Bertsimas et al., 2011] Dimitris Bertsimas, David B. Brown, and Constantine Caramanis. Theory and applications of robust optimization. SIAM Review, 53:464-501, 2011.","title":"Theory and applications of robust optimization","authors":"Dimitris Bertsimas, David B Brown, Constantine Caramanis","year":"2011","venue":"SIAM Review","doi":null,"url":null,"internal_url":null},{"position":6,"raw":"[Cheng, 1984] Philip E. Cheng. Strong consistency of near- est neighbor regression function estimators. Journal of Multivariate Analysis, 15:63-72, 1984.","title":"Strong consistency of nearest neighbor regression function estimators","authors":"Philip E Cheng","year":"1984","venue":"Journal of Multivariate Analysis","doi":null,"url":null,"internal_url":null},{"position":7,"raw":"[Elmachtoub and Grigas, 2022] Adam N. Elmachtoub and Paul Grigas. Smart \"predict, then optimize\". Management Science, 68:9-26, 2022.","title":"Smart \"predict, then optimize","authors":"Adam N Elmachtoub, Paul Grigas","year":"2022","venue":"Management Science","doi":null,"url":null,"internal_url":null},{"position":8,"raw":"[Ferber et al., 2020] Aaron Ferber, Bryan Wilder, Bistra Dilk- ina, and Milind Tambe. MIPaaL: Mixed integer program as a layer. Proceedings of the AAAI Conference on Artificial Intelligence, 34:1504-1511, 2020.","title":"MIPaaL: Mixed integer program as a layer","authors":"Aaron Ferber, Bryan Wilder, Bistra Dilkina, Milind Tambe","year":"2020","venue":"Proceedings of the AAAI Conference on Artificial Intelligence","doi":null,"url":null,"internal_url":null},{"position":9,"raw":"[Geman et al., 1992] Stuart Geman, Elie Bienenstock, and Renc Doursat. Neural networks and the bias/variance dilemma. Neural computation, 4:1-58, 1992.","title":"Neural networks and the bias/variance dilemma","authors":"Stuart Geman, Elie Bienenstock, Renc Doursat","year":"1992","venue":"Neural computation","doi":null,"url":null,"internal_url":null},{"position":10,"raw":"[Gurobi Optimization, 2023] Gurobi Optimization. Gurobi Optimizer 10.0.1 reference manual, 2023. Accessed: 2024- 04-23.","title":"Gurobi Optimization","authors":null,"year":"2023","venue":"Gurobi Optimizer","doi":null,"url":null,"internal_url":null},{"position":11,"raw":"[Ifrim et al., 2012] Georgiana Ifrim, Barry O'Sullivan, and Helmut Simonis. Properties of energy-price forecasts for scheduling. International Conference on Principles and Practice of Constraint Programming, pages 957-972, 2012.","title":"Properties of energy-price forecasts for scheduling","authors":"Georgiana Ifrim, O' Barry, Helmut Sullivan, Simonis","year":"2012","venue":"International Conference on Principles and Practice of Constraint Programming","doi":null,"url":null,"internal_url":null},{"position":12,"raw":"[Jiang, 2019] Heinrich Jiang. Non-asymptotic uniform rates of consistency for k-nn regression. Proceedings of the AAAI Conference on Artificial Intelligence, 33:3999-4006, 2019.","title":"Non-asymptotic uniform rates of consistency for k-nn regression","authors":"Heinrich Jiang","year":"2019","venue":"Proceedings of the AAAI Conference on Artificial Intelligence","doi":null,"url":null,"internal_url":null},{"position":13,"raw":"[Kiureghian and Ditlevsen, 2009] Armen Der Kiureghian and Ove Ditlevsen. Aleatory or epistemic? Does it matter? Structural Safety, 31:105-112, 2009.","title":"Armen Der Kiureghian and Ove Ditlevsen. Aleatory or epistemic? Does it matter?","authors":"Ditlevsen Kiureghian","year":"2009","venue":"Structural Safety","doi":null,"url":null,"internal_url":null},{"position":14,"raw":"[Kukačka et al., 2017] Jan Kukačka, Vladimir Golkov, and Daniel Cremers. Regularization for deep learning: A tax- onomy. arXiv preprint arXiv:1710.10686, 2017.","title":"Regularization for deep learning: A taxonomy","authors":"Kukačka","year":"2017","venue":null,"doi":null,"url":null,"internal_url":null},{"position":15,"raw":"[Mandi and Guns, 2020] Jayanta Mandi and Tias Guns. Inte- rior point solving for LP-based prediction+optimisation. Advances in Neural Information Processing Systems, 33:7272-7282, 2020.","title":"Jayanta Mandi and Tias Guns. Interior point solving for LP-based prediction+optimisation","authors":"Guns Mandi","year":"2020","venue":"Advances in Neural Information Processing Systems","doi":null,"url":null,"internal_url":null},{"position":16,"raw":"[Mandi et al., 2022] Jayanta Mandi, Víctor Bucarey, Maxime Mulamba, and Tias Guns. Decision-focused learning: Through the lens of learning to rank. International Con- ference on Machine Learning, PMLR, 162:14935-14947, 2022.","title":"Decision-focused learning: Through the lens of learning to rank","authors":"Jayanta Mandi, Víctor Bucarey, Maxime Mulamba, Tias Guns","year":"2022","venue":"International Conference on Machine Learning","doi":null,"url":null,"internal_url":null},{"position":17,"raw":"[Mulamba et al., 2021] Maxime Mulamba, Jayanta Mandi, Michelangelo Diligenti, Michele Lombardi, Victor Bucarey, and Tias Guns. Contrastive losses and solution caching for predict-and-optimize. Proceedings of the Thirtieth Interna- tional Joint Conference on Artificial Intelligence, IJCAI-21, pages 2833-2840, 2021.","title":"Contrastive losses and solution caching for predict-and-optimize","authors":"Maxime Mulamba, Jayanta Mandi, Michelangelo Diligenti, Michele Lombardi, Victor Bucarey, Tias Guns","year":"2021","venue":"Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, IJCAI-21","doi":null,"url":null,"internal_url":"/proceedings/2021/390"},{"position":18,"raw":"[Sahoo et al., 2023] Subham Sekhar Sahoo, Anselm Paulus, Marin Vlastelica, Vít Musil, Volodymyr Kuleshov, and Georg Martius. Backpropagation through combinatorial algorithms: Identity with projection works. The Eleventh In- ternational Conference on Learning Representations, 2023.","title":"Backpropagation through combinatorial algorithms: Identity with projection works","authors":"Sekhar Subham, Anselm Sahoo, Marin Paulus, Vít Vlastelica, Volodymyr Musil, Georg Kuleshov, Martius","year":"2023","venue":"The Eleventh International Conference on Learning Representations","doi":null,"url":null,"internal_url":null},{"position":19,"raw":"[Shah et al., 2022] Sanket Shah, Kai Wang, Bryan Wilder, Andrew Perrault, and Milind Tambe. Decision-focused learning without decision-making: Learning locally opti- mized decision losses. Advances in Neural Information Processing Systems, 35:1320-1332, 2022.","title":"Decision-focused learning without decision-making: Learning locally optimized decision losses","authors":"Sanket Shah, Kai Wang, Bryan Wilder, Andrew Perrault, Milind Tambe","year":"2022","venue":"Advances in Neural Information Processing Systems","doi":null,"url":null,"internal_url":null},{"position":20,"raw":"[Silvestri et al., 2023] Mattia Silvestri, Senne Berden, Jayanta Mandi, Ali İrfan Mahmutogulları, Maxime Mulamba, Allegra De Filippo, Tias Guns, and Michele Lombardi. Score function gradient estimation to widen the applicability of decision-focused learning. arXiv preprint arXiv:2307.05213, 2023.","title":"Score function gradient estimation to widen the applicability of decision-focused learning","authors":"Mattia Silvestri, Senne Berden, Jayanta Mandi, Ali İrfan Mahmutogulları, Maxime Mulamba, Allegra De Filippo, Tias Guns, Michele Lombardi","year":"2023","venue":null,"doi":null,"url":null,"internal_url":null},{"position":21,"raw":"[Simonis et al., 1999] Helmut Simonis, Barry O'Sullivan, Deepak Mehta, Barry Hurley, and Milan De Cauwer. CSPLib problem 059: Energy-cost aware scheduling. http://www.csplib.org/Problems/prob059, 1999. Accessed: 2024-04-23.","title":"CSPLib problem 059: Energy-cost aware scheduling","authors":"Helmut Simonis, O' Barry, Deepak Sullivan, Barry Mehta, Milan De Hurley, Cauwer","year":"2024","venue":null,"doi":null,"url":"http://www.csplib.org/Problems/prob059,1999","internal_url":null},{"position":22,"raw":"[Tang and Khalil, 2022] Bo Tang and Elias B. Khalil. PyEPO: A PyTorch-based end-to-end predict-then-optimize library for linear and integer programming. arXiv preprint arXiv:2206.14234, 2022.","title":"PyEPO: A PyTorch-based end-to-end predict-then-optimize library for linear and integer programming","authors":"Bo Tang, Elias B Khalil","year":"2022","venue":null,"doi":null,"url":null,"internal_url":null},{"position":23,"raw":"[Veviurko et al., 2024] Grigorii Veviurko, Wendelin Böhmer, and Mathijs de Weerdt. You shall pass: Dealing with the zero-gradient problem in predict and optimize for convex optimization. arXiv preprint arXiv:2307.16304, 2024.","title":"You shall pass: Dealing with the zero-gradient problem in predict and optimize for convex optimization","authors":"Grigorii Veviurko, Wendelin Böhmer, Mathijs De, Weerdt","year":"2024","venue":null,"doi":null,"url":null,"internal_url":null},{"position":24,"raw":"[Vlastelica et al., 2020] Marin Vlastelica, Anselm Paulus, Vít Musil, Georg Martius, and Michal Rolínek. Differ- entiation of blackbox combinatorial solvers. International Conference on Learning Representations, 2020.","title":"Differentiation of blackbox combinatorial solvers","authors":"Marin Vlastelica, Anselm Paulus, Vít Musil, Georg Martius, Michal Rolínek","year":"2020","venue":"International Conference on Learning Representations","doi":null,"url":null,"internal_url":null},{"position":25,"raw":"[Wilder et al., 2019] Bryan Wilder, Bistra Dilkina, and Milind Tambe. Melding the data-decisions pipeline: Decision-focused learning for combinatorial optimization. Proceedings of the AAAI Conference on Artificial Intelli- gence, 33:1658-1665, 2019.","title":"Melding the data-decisions pipeline: Decision-focused learning for combinatorial optimization","authors":"Bryan Wilder, Bistra Dilkina, Milind Tambe","year":"2019","venue":"Proceedings of the AAAI Conference on Artificial Intelligence","doi":null,"url":null,"internal_url":null}],"license_note":"Copyright International Joint Conferences on Artificial Intelligence."}