The performance of heuristic optimization methods is strongly influenced by how problems are modeled and parameterized. Despite this, the interplay between problem formulation and solver behavior remains underexplored in comparison to algorithmic design itself. This special session aims to bridge that gap by highlighting the mutual dependencies between models and solvers, and by fostering discussions on how to systematize problem modeling for heuristic optimization.

Several challenges motivate this session. The same problem can often be expressed in multiple ways, and these formulations may interact very differently with different search paradigms: constructive, local search-based, or other metaheuristics. Unlike exact methods, where established modeling paradigms (e.g., MILP formulations) exist, heuristic optimization still lacks agreed-upon standards for representing problems and interfacing with solvers. This often requires significant tailoring by practitioners, particularly in industry, where solvers are frequently adjusted to fit specific models or, conversely, models are reformulated to accommodate available solvers.

Topics of interest

We invite contributions that advance the understanding of model-solver interactions and explore ways to formalize and standardize this process. Relevant topics include:

  • Model-solver interaction in heuristic optimization
  • Real-world case studies highlighting modeling choices and their effects on solver performance
  • High and low-level landscape features extraction for problem/model analysis
  • Parametrization, proxy functions, and multi-fidelity problem formulations
  • Efforts toward standardizing how problem models and solvers communicate
  • Benchmark design and the biases introduced by modeling choices
  • Algorithmic developments inspired by specific modeling conditions or problem features
  • Online usage of solver feedback to inform better modeling choices
  • Best practices and methodologies for effective heuristic modeling

We particularly encourage submissions that combine practical insights from real-world applications with methodological contributions. The goal is to create a space where researchers can exchange experiences, identify common challenges, and move toward a more systematic and standardized approach to modeling for heuristics.

Organisers

  • Elena Raponi
    Leiden University
    e.raponi(at)liacs.leidenuniv.nl
  • Carlos M. Fonseca
    University of Coimbra
    cmfonsec(at)dei.uc.pt