Home

Awesome

awesome-moo-ml-papers

Pareto set learning

  1. PHN: Learning Pareto front by hypernetwork
    Authors: Navon et al
    Conference: ICLR, 2019
    Link: arXiv:2010.04104

  2. PHN-HVI: Improving pareto front learning via multi-sample hyper-networks
    Author: LP Hoang et al
    Conference: AAAI, 2023
    Link: arXiv:2212.01130

  3. PSL-Frame: A framework for controllable pareto front learning with completed scalarization functions and its applications
    Author: TA Tuan et al
    Journal: Neural Networks, 2024
    Link: arXiv:2302.12487

  4. PSL-Exp: Pareto set learning for expensive multiobjective optimization
    Authors: Lin et al
    Conference: NeurIPS, 2022
    Link: arXiv:2210.08495

  5. COSMOS: Scalable Pareto Front Approximation for Deep Multi-Objective Learning Authors: Ruchte et al
    Conference: ICDM, 2022 Link: arXiv:2103.13392.pdf

  6. PaMaL: Pareto manifold learning: Tackling multiple tasks via ensembles of single-task models
    Authors: Dimitriadis et al
    Conference: ICLR, 2023 Link: Proceedings of Machine Learning Research (PMLR)

  7. GMOOAR: Multi-objective deep learning with adaptive reference vectors Authors: Ruchte et al Conference: NeurIPS, 2023 Link: NeurIPS Conference Paper

  8. HVPSL: Multi-objective deep learning with adaptive reference vectors Authors: Xiaoyuan Zhang et al
    Conference: NeurIPS, 2023 Link: NeurIPS Conference Paper

  9. Smooth Tchebycheff Scalarization for Multi-Objective Optimization Authors: Xi Lin et al
    Conference: ICML 2024
    Link: arxiv

  10. Low Rank (LoRA) PSL.

  11. Learning a Neural Pareto Manifold Extractor with Constraints
    Authors: Gupta et al
    Conference: UAI 2021

Pareto Multitask learning (Discrete Solutions)

  1. PMTL: Pareto Multi Task Learning. NeurIPS, 2018.

  2. MGDA. (Sener 2018).

  3. EPO.

  4. MOO-SVGD Profiling Pareto Front With Multi-Objective Stein Variational Gradient Descent. Conference: NeurIPS 2022. Author: Xingchao Liu

  5. GMOOAR. Multi-objective deep learning with adaptive reference vectors.

  6. PNG.

Using MOO idea to solve MTL only for a single solution

It is noticed that, this line using Pareto ideas to solve MTL.

  1. Nash-MTL. Navon 2022.

Theories.

  1. HVPSL: Multi-objective deep learning with adaptive reference vectors Authors: Xiaoyuan Zhang et al
    Conference: NeurIPS, 2023 Link: NeurIPS Conference Paper TL: Understanding the generlization bound of PSL.

  2. Revisiting scalarization in multi-task learning: A theoretical perspective Authors: Yuzheng Hu et al
    Conference: NeurIPS, 2023 Link: NeurIPS Conference Paper TL: When MOO-MTL actually has no tradeoff.

Applications in very large problems.

A. Drug deign

B. LLM

  1. Panacea: Pareto Alignment via Preference Adaptation for LLMs Authors: Yifan Zhong et al
    Conference: Unknown
    Link: arxiv

  2. Controllable Preference Optimization.

NN meets MOEA.

  1. Pseudo Weight Net: Learning to Predict Pareto-optimal Solutions From Pseudo-weights Author: Deb. Jornal: TEVC link: https://www.egr.msu.edu/~kdeb/papers/c2022010.pdf

Awesome MOO libs

  1. libMTL. Yu Zhang's group. Sustech.

  2. Libmoon. Xiaoyuan Zhang. CityU HK.