[NeurIPS 2024] ReEvo: Large Language Models as Hyper-Heuristics with Reflective Evolution

[AAAI-25] HSEvo: Elevating Automatic Heuristic Design with Diversity-Driven Harmony Search and Genetic Algorithm Using LLMs

Some hyper-heurisics from CHeSC 2011 challenge and the challenge results reproducing attempts.

[AAAI-26] MPaGE: Pareto-Grid-Guided Large Language Models for Fast and High-Quality Heuristics Design in Multi-Objective Combinatorial Optimization

Online selection hyper-heuristic with generic parameter control in low-level heuristics (meta-heuristic).

Training Feedforward Neural Networks with Bayesian Hyper-Heuristics

Multi-Objective Agent-Based Hyper-Heuristic

HyFlex-compatible implementation of the UAV Zoo Feeding Problem (UZF) with a selection hyper-heuristic for solving combinatorial optimisation instances.

Thompson Sampling HH implementation to solve TSP in a genetic algorithm configuration.

Implementing hyper-heuristic selection strategies towards creating a synergy between them.

HyPy is a general hyper-heuristic package for solving combinatorial optimization problems by employing and developing hyper-heuristics.

This is my Python implementation of a Hybrid Perturbative Hyper-Heuristic (HPHH) to optimize 24 benchmark mathematical functions.

Water Distribution Network simulator using EPANET

Hybrid classical-quantum metaheuristic framework for identical parallel machine scheduling (P||Cmax). Encodes local subproblems as QUBO and solves with shallow QAOA (Qiskit). A UCB1 multi-armed bandit hyperheuristic dynamically selects between quantum and classical operators to minimize makespan.

A Matlab-based Hyper-Heuristic framework. To cite this Original Software Publication: https://www.sciencedirect.com/science/article/pii/S2352711022000413

JDHS is a java library for dynamic heuristic sets that can be used within hyper-heuristics

Reinforcement learning hyper-heuristic for the identical parallel machine scheduling problem (P||Cmax). A DQN agent dynamically selects low-level scheduling operators (swap, relocate, balance) to minimize makespan. Includes LPT/list scheduling baselines, Gymnasium env, and benchmarking pipeline.

A Hyflex-compatible problem domain with dynamic feasibility calculation and a realistic problem instance generator