Algorithms

solvOR packs 40+ algorithms into one library. Here's how to find the right one.

By Category

Category	Solvers	When to Use
Linear Programming	solve_lp, solve_milp	Linear objectives, linear constraints
Constraint Programming	solve_sat, Model, solve_exact_cover	Logic puzzles, satisfiability, exact cover
Metaheuristics	anneal, tabu_search, lns, evolve	Combinatorial, no gradients
Continuous	adam, bfgs, nelder_mead, bayesian_opt	Smooth functions, gradients or black-box
Graph	dijkstra, astar, max_flow, kruskal	Paths, flows, spanning trees
Combinatorial	solve_knapsack, solve_bin_pack, solve_job_shop	Packing, scheduling, routing

Quick Decision Tree

Is your objective linear?
├─ Yes → Do you have integer variables?
│        ├─ Yes → solve_milp
│        └─ No → solve_lp
│
└─ No → Can you compute gradients?
         ├─ Yes → Is evaluation expensive?
         │        ├─ Yes → bfgs or lbfgs
         │        └─ No → adam
         │
         └─ No → Is the problem discrete/combinatorial?
                  ├─ Yes → Is it a puzzle/exact cover?
                  │        ├─ Yes → solve_exact_cover or Model
                  │        └─ No → anneal or tabu_search
                  │
                  └─ No → Is evaluation expensive?
                           ├─ Yes → bayesian_opt
                           └─ No → nelder_mead or differential_evolution

Guarantees

Some solvers prove optimality, others don't.

Guarantee	Solvers
Optimal	solve_lp, solve_milp, dijkstra, bellman_ford, kruskal, solve_hungarian
Heuristic	anneal, tabu_search, evolve, particle_swarm
Local optimum	gradient_descent, adam, bfgs, nelder_mead

When you need a certificate of optimality, use the exact solvers. When you need speed and "good enough," use heuristics.