SAS^® Optimization Features

Algebraic, symbolic optimization modeling language 

• Flexible algebraic syntax for intuitive model formulation.
• Support for the transparent use of SAS functions.
• Direct invocation of linear, nonlinear, quadratic and mixed-integer solvers.
• Support for the rapid prototyping of customized optimization algorithms, including support for named problems and subproblems.
• Use of industry-standard MPS/QPS format input data sets.
• Aggressive presolvers to reduce effective problem size.
• LP (linear programming) and MILP (mixed integer linear programming) solvers deliver improvements in performance, shortening the time needed to reach optimality and enabling you to solve more complex problems in a given amount of time.

Powerful optimization solvers 

• Linear programming solution algorithms: 
  • Primal and dual simplex.
  • Network simplex.
  • Interior point with (experimental) crossover.
  • Concurrent solve capability.
• Mixed integer linear programming solution algorithms: 
  • Branch-and-bound integer with cutting planes.
  • Primal heuristics.
  • Conflict search.
  • Option tuning.
  • Added control on the solution of the root node (LP relaxation) problem.
  • Report up to the K best integer feasible solutions or up to K optimal solutions.
• Decomposition algorithm (automated Dantzig-Wolfe) for linear programming and mixed-integer linear programming problems with block-angular, block-diagonal or embedded network structure.
• Quadratic programming solution algorithm: interior point with state-of-the-art solver tailored for large-scale optimization problems.
• Nonlinear programming solution algorithms: active set, interior point. Concurrent solve capability. Multistart algorithm for nonconvex problems.

Network optimization

• Optimization and diagnostic algorithms include:
  • Connected components and biconnected components (with articulation points).
  • Maximal clique enumeration.
  • Cycle enumeration.
  • Transitive closure.
  • Minimum cut.
  • Minimum spanning tree.
  • Summary statistics.
  • Minimal-cost linear assignment.
  • Minimum-cost network flow.
  • Shortest path.
  • Traveling salesman problem.
  • Path enumeration.
  • Multiple links between pairs of nodes can be input and processed.

Black-box optimization

• Solves problems with nonlinear functions that can be non-smooth, discontinuous, not continuously differentiable, and so on.
• Hybrid parallel algorithm, including generic algorithms, global GA-type heuristics and pattern search. Multiobjective optimization.

Constraint programming

• Solves constraint satisfaction problems using finite-domain constraint programming, with domain reduction/constraint propagation and a choice of search strategies (look ahead and backtracking). Find one, several or all feasible solutions. Optionally specify an objective function and find an optimal solution (bisection search method).

Distributed, accessible & cloud-ready

• Optimization solvers run on SAS Viya, a scalable and distributed in-memory analytics platform.
• Distributes analysis and data tasks across multiple computing nodes.
• Distributed computation features:
  • Multistart option for nonlinear (NLP) solver in PROC OPTMODEL.
  • Decomposition algorithm (MILP) in PROC OPTMODEL, PROC OPTMILP.
  • Solving independent optimization scenarios: COFOR loop in PROC OPTMODEL.
  • Concurrent mode for MILP solver (PROC OPTMODEL, PROC OPTMILP).
  • Branch-and-bound MILP solver algorithm (PROC OPTMODEL, PROC OPTMILP).
  • Local search optimization in PROC OPTMODEL.
  • Shortest path and connected components network algorithms in PROC OPTNETWORK.
  • BY-group processing in network algorithms in PROC OPTNETWORK.
  • Provides fast, concurrent, multiuser access to data in memory.
  • Includes fault tolerance for high availability.
  • Lets you add the power of SAS Analytics to other applications using SAS Viya REST APIs.