Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Videos

Deep Reinforcement Learning for Multiobjective Optimization.

Kaiwen Li, Tao Zhang, Rui Wang

IEEE Transactions on Cybernetics

|March 20, 2020

Summary

This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

A receptor-like mechanosensitive protein governs preprophase band positioning for asymmetric cell divisions and SC morphogenesis.

Proceedings of the National Academy of Sciences of the United States of America·2026

Same author

Comprehensive insights into the roles of eosinophils in the mucosal immunity of gastrointestinal tract: defenders, destructors, and modulators.

Frontiers in immunology·2026

Same author

Bridging Information Asymmetry: A Hierarchical Framework for Deterministic Blind Face Restoration.

IEEE transactions on pattern analysis and machine intelligence·2026

Same author

Dual-Affinity Interphase Engineering Enables Stable Aqueous Zn-S Batteries.

Angewandte Chemie (International ed. in English)·2026

Same author

Vertical farming as a land sparing strategy: GHG implications for UK agricultural landscapes.

Cleaner food systems·2026

Same author

Ecological conservation priority requires foresight: coupling multiple scenarios and network perspective.

Journal of environmental management·2026

Same journal

An Evolutionary Algorithm Assisted by an Ensemble of Pareto-Optimal Surrogate Models.

IEEE transactions on cybernetics·2026

Same journal

A Quantum Self-Attention Neural Network Model on Quantum Circuits.

IEEE transactions on cybernetics·2026

Same journal

Semi-Explicit Solution of Some Discrete-Time Higher-Order-Cost Mean-Field-Type Control.

IEEE transactions on cybernetics·2026

Same journal

A Novel One-Step Small Object Detector for Autonomous Aerial Vehicles.

IEEE transactions on cybernetics·2026

Same journal

Online Data-Driven-Based Optimal Output Tracking Control Without Initial Stabilizing Policy.

IEEE transactions on cybernetics·2026

Same journal

Digital Redesign-Based Interval State Estimation for Continuous Systems With Aperiodic Discrete Measurements.

IEEE transactions on cybernetics·2026

See all related articles

This study introduces a deep reinforcement learning (DRL) framework for multiobjective optimization problems (MOPs). The DRL-based multiobjective optimization algorithm (DRL-MOA) offers fast, generalizable solutions without retraining.

Area of Science:

Artificial Intelligence
Operations Research
Computational Science

Background:

Multiobjective optimization problems (MOPs) are complex, requiring efficient solution methods.
Existing methods often struggle with scalability and generalization for MOPs.

Purpose of the Study:

To propose a novel end-to-end framework, DRL-based multiobjective optimization algorithm (DRL-MOA), for solving MOPs.
To leverage deep reinforcement learning (DRL) for direct Pareto-optimal solution generation.

Main Methods:

Decomposition of MOPs into scalar subproblems, each modeled as a neural network.
Collaborative optimization of subproblem parameters using a neighborhood-based parameter-transfer strategy and DRL.
Application to the multiobjective traveling salesman problem (MOTSP) using Pointer Networks.

Related Experiment Videos

Main Results:

DRL-MOA achieves fast solving times through direct neural network forward calculation, eliminating iterative processes.
Trained models exhibit strong generalization ability, scaling to new problems without retraining.
Experimental results demonstrate effectiveness and competitiveness against benchmark methods in performance and speed.

Conclusions:

DRL-MOA provides a novel, efficient approach to MOPs using DRL.
The method offers significant advantages in generalization and solving speed compared to existing techniques.
DRL-MOA presents a promising new direction for tackling complex optimization challenges.