R2024b

New Features, Bug Fixes, Compatibility Considerations

 Detailed plot function for fmincon

The fmincon optimplot plot function displays detailed information during an optimization. For example,

x0 = [-2;2];
fun = @(x)100*(x(2) - x(1)^2)^2 + (1 - x(1))^2;
lb = [-2;-2];
ub = [3;3];
options = optimoptions("fmincon",PlotFcn="optimplot");
[x,fval] = fmincon(fun,x0,[],[],[],[],lb,ub,[],options)

The optimplot plot function opens in a separate window from any other plot functions. For an example, see Monitor Solution Process with optimplot.

  Plot Functions and Output Functions for intlinprog "highs" Algorithm: Obtain iterative information as the solver progresses

The intlinprog "highs" algorithm now supports output functions and plot functions. For example, to have intlinprog produce a plot while solving, specify the built-in "optimplotmilp" plot function.

options = optimoptions("intlinprog",PlotFcn="optimplotmilp");
[x,fval] = intlinprog(f,intcon,A,b,Aeq,beq,lb,ub,options)

For details, see intlinprog Output Function and Plot Function Syntax.

 Compatibility Considerations

The intlinprog "legacy" algorithm will be removed in a future release.

The optimValues output function and plot function structure field named lowerbound is now named dualbound.

The "highs" algorithm now supports the MaxFeasiblePoints option. The output structure for the "highs" algorithm now supports the numfeaspoints field.

 Code Generation for coneprog: Generate C or C++ code for problems with linear objective function and cone constraints

The coneprog function now supports code generation. For details, see Code Generation for coneprog Background and Generate Code for coneprog.

Single-precision code generation for quadprog, lsqlin, and optimwarmstart

The quadprog and lsqlin solvers and the optimwarmstart function can generate code for single-precision floating point hardware. If you want to generate single-precision code, all relevant data, such as constraint matrices and bounds, must have the type "single". For details, see Single-Precision Code Generation.

Note

This single-precision capability is available for code generation only; it is not available for general MATLAB^® computations.

EquationProblem and OptimizationConstraint Evaluation: Compute equation and constraint satisfaction using issatisfied and evaluate

Compute the degree of satisfaction of nonlinear equation systems and of optimization constraints using the evaluate and issatisfied functions. For an EquationProblem object, issatisfied returns a boolean value indicating whether all equations are satisfied at a point, and returns a boolean satisfaction value for each equation in a system in the second output. evaluate returns the equation value at a point. For an OptimizationConstraint object, evaluate returns the values of the constraints, and issatisfied returns boolean values for the constraints.

The value of a constraint depends on the constraint type. An equation is equivalent to an == constraint. For expressions L and R:

Constraint Type	Value
L <= R	L – R
L >= R	R – L
L == R	abs(L – R)

For details, see the evaluate and issatisfied reference pages.

 mpsread uses HiGHS software

mpsread now uses HiGHS open-source software to read MPS files into a form that intlinprog and linprog accept.

 Compatibility Considerations

The mpsread display now includes a reference to HiGHS.

Green hydrogen production example

The example Optimize Green Hydrogen Production System shows how to use forecast cost and demand data to plan a hydrogen production system optimally. The example uses the problem-based optimization approach to model and solve a large linear programming problem.

 Functionality being removed or changed

intlinprog "legacy" algorithm and Algorithm option will be removed in a future release

Still runs

The intlinprog "legacy" algorithm will be removed in a future release. Usually, the "highs" algorithm (default) performs better than the "legacy" algorithm. Now that "highs" supports output functions and plot functions, there is no loss of functionality when switching to "highs".

Because intlinprog will have just one algorithm when the "legacy" algorithm is removed, at that time:

• The Algorithm option will be removed.

• All options that are valid only for the "legacy" algorithm will be removed.

• intlinprog will use the "highs" algorithm, the current default.

• The algorithm field will be removed from the output structure.

intlinprog plot function and output function optimValues field lowerbound renamed dualbound

Behavior change

The optimValues output function and plot function structure field named lowerbound is now named dualbound. The new name is clearer for maximization problems, where the dualbound is an upper bound on the objective function value. The optimplotmilp plot function now uses the labels Primal Bound and Dual Bound instead of Branch/bound UB and Branch/bound LB.

To avoid errors, change any custom output functions and plot functions to use the new field name.

linprog "dual-simplex-legacy" algorithm will be removed in a future release

Still runs

The linprog "dual-simplex-legacy" algorithm will be removed in a future release. Usually, the "dual-simplex-highs" algorithm (default) performs better than the "dual-simplex-legacy" algorithm.

More objects support evaluate and issatisfied

The following objects now support the evaluate and issatisfied functions:

• EquationProblem

• OptimizationConstraint

• OptimizationEquality

• OptimizationInequality

You can evaluate the values of these objects (optimization expressions or constraints or equations) on a set of points that are expressed as an OptimizationValues object or on a single point expressed as a structure. For an example, see Evaluate Expressions in Equation.