Comparer des méthodes

	Matheuristics	Optimisation stochastique
Domaine	Optimisation	Optimisation
Famille	Process / pipeline	Process / pipeline
Année d'origine	2009	1951 (SGD); 2014 (Adam)
Auteur d'origine	Maniezzo, Stützle & Voß	—
Type	Hybrid optimization framework	Gradient-based iterative optimization
Source fondatrice	Maniezzo, V., Stützle, T., & Voß, S. (Eds.). (2009). Matheuristics: Hybridizing Metaheuristics and Mathematical Programming. Springer. ISBN: 978-1-4419-1305-0	Robbins, H. & Monro, S. (1951). A Stochastic Approximation Method. Annals of Mathematical Statistics, 22(3), 400-407.
Alias	Hybrid Metaheuristics, MIP-based Heuristics, Math-Programming Hybrids, Matematiksel Sezgisel Yöntemler	Stokastik Optimizasyon (SGD & Varyantları), stochastic gradient descent, SGD, Adam
Apparentées	3	3
Résumé	Matheuristics is a class of hybrid optimization methods that tightly couple exact mathematical programming components—such as mixed-integer programming (MIP) solvers—with metaheuristic search procedures. Formally introduced and named by Maniezzo, Stützle, and Voß in 2009, the framework leverages the global-search capability of metaheuristics and the structural exploitation of mathematical programming to tackle large-scale combinatorial optimization problems that neither approach can solve effectively alone.	Stochastic optimization is a family of iterative methods that minimize an objective function by computing gradients on randomly sampled subsets of data — mini-batches — rather than on the entire dataset at once. Pioneered by Robbins and Monro in 1951 as stochastic approximation, the approach became the standard engine for training large-scale machine-learning models through variants such as SGD with momentum, AdaGrad, RMSProp, and Adam.