Classical algorithms

These pages cover the implemented solver wrappers, NumPy searches, and simple reference baselines. Read the shared objectives and inputs first. Entries without a study method ID remain available through the Python API.

Exact references and optimization backends

• Exhaustive binary search

• Markowitz support enumeration

• Convex Markowitz QP with OSQP

• Binary MILP with HiGHS

• Binary quadratic optimization with SCIP

Search and relaxation methods

• QP relaxation, rounding, and refinement

• Multistart swap local search

• Simulated annealing

• Parallel tempering / replica exchange

• Tabu search

• Memetic genetic algorithm

• Iterated local search

• GRASP

• Variable neighborhood search

• Projected gradient descent

• Frank–Wolfe with away steps

Simple comparison baselines

• Top-return selection

• Minimum-variance binary selection

• Equal-weight allocation

• Random exact-K sampling