mlsanityData to 14 Oct 2026

Adaptive Online Mirror Descent for Tchebycheff Scalarization in Multi-Objective Learning

arXiv 2410.21764 ↗First posted 29 Oct 2024cs.LG cs.AIcited by 4

Not on the map. Why?

Abstract (arXiv)

Multi-objective learning (MOL) aims to learn under multiple potentially conflicting objectives and strike a proper balance. While recent preference-guided MOL methods often rely on additional optimization objectives or constraints, we consider the classic Tchebycheff scalarization (TCH) that naturally allows for locating solutions with user-specified trade-offs. Due to its minimax formulation, directly optimizing TCH often leads to training oscillation and stagnation. In light of this limitation, we propose an adaptive online mirror descent algorithm for TCH, called (Ada)OMD-TCH. One of our main ingredients is an adaptive online-to-batch conversion that significantly improves solution optimality over traditional conversion in practice while maintaining the same theoretical convergence guarantees. We show that (Ada)OMD-TCH achieves a convergence rate of mathcal O(√log m/T), where m is the number of objectives and T is the number of rounds, providing a tighter dependency on m in the offline setting compared to existing work. Empirically, we demonstrate on both synthetic problems and federated learning tasks that (Ada)OMD-TCH effectively smooths the training process and yields preference-guided, specific, diverse, and fair solutions.

Key facts. Cited by 4 papers in the map (data through 14 Oct 2026). Builds on 26 of its 57 references that are in the map. First posted on arXiv 29 Oct 2024.

Builds on

26 of 57 references.

References · 26

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2016Communication-Efficient Learning of Deep Networks from Decentralized Datacited by 7,669
  3. 2021Convex optimizationcited by 2,515
  4. 2020Gradient Surgery for Multi-Task Learningcited by 1,256
  5. 2017GradNorm: Gradient Normalization for Adaptive Loss Balancing in Deep Multitask Networkscited by 834
  6. 2018Multi-Task Learning as Multi-Objective Optimizationcited by 821
  7. 2019Introduction to Online Convex Optimizationcited by 803
  8. 2021Exploiting Shared Representations for Personalized Federated Learningcited by 541
  9. 2020An Efficient Framework for Clustered Federated Learningcited by 511
  10. 2021Conflict-Averse Gradient Descent for Multi-task Learningcited by 459
  11. 2019Agnostic Federated Learningcited by 456
  12. 2019Fair Resource Allocation in Federated Learningcited by 394
14 more
  1. 2019Pareto Multi-Task Learningcited by 224
  2. 2021MARS: Markov Molecular Sampling for Multi-objective Drug Discoverycited by 125
  3. 2020Tilted Empirical Risk Minimizationcited by 106
  4. 2021Federated Learning with Fair Averagingcited by 61
  5. 2024Smooth Tchebycheff Scalarization for Multi-Objective Optimizationcited by 54
  6. 2020Federated Learning Meets Multi-objective Optimizationcited by 53
  7. 2022Pareto Set Learning for Expensive Multi-Objective Optimizationcited by 51
  8. 2023Revisiting Scalarization in Multi-Task Learning: A Theoretical Perspectivecited by 51
  9. 2023Three-Way Trade-Off in Multi-Objective Learning: Optimization, Generalization and Conflict-Avoidancecited by 36
  10. 2024Robust Multi-Task Learning with Excess Riskscited by 24
  11. 2022Proportional Fairness in Federated Learningcited by 21
  12. 2024LibMOON: A Gradient-based MultiObjective OptimizatioN Library in PyTorchcited by 10
  13. 2021Pareto Efficient Fairness in Supervised Learning: From Extraction to Tracingcited by 8
  14. 2021Exact Pareto Optimal Search for Multi-Task Learning and Multi-Criteria Decision-Makingcited by 6

Built on by

Cited by year

Citing papers per year: 2025: 4.

4 papers.

  1. 2025Gradient-Based Multi-Objective Deep Learning: Algorithms, Theories, Applications, and Beyondcited by 18
  2. 2025Multi-objective Large Language Model Alignment with Hierarchical Expertscited by 13
  3. 2025AReUReDi: Annealed Rectified Updates for Refining Discrete Flows with Multi-Objective Guidancecited by 5
  4. 2025AutoScale: Linear Scalarization Guided by Multi-Task Optimization Metricscited by 2