|
|
|
Added a nonlinear operator
to overload the concave function sqrt . Do NOT use
this command in problems you intend to solve the problem using geometric
programming or a non-conic solver.
|
|
Added a nonlinear operator
pwf
to model piece-wise functions (experimental functionality serving as a good example of advanced use
of nonlinear operators.)
|
|
|
|
Cleaned up the engine behind the convexity analysis of nonlinear operators
(Should be more robust and less conservative, but might be a bit
slower now. Please report any performance issues.)
|
|
The function detinv has been removed (use
geomean instead) |
|
Changed format for specification of nonlinear operators.
If you have created your own, make sure to update it according
to the tracenorm
example (second output needs to be changed)
|
|
|
Resolved a bug introduced in R20050720 causing symmetry
reduction of sum of squares problem to be skipped in some cases.
|
|
Resolved a bug which caused x'*x to act as x.'*x on complex valued column-vectors x.
|
|
|
Fixed a rather serious bug affecting
double
on nonlinear variables (bug introduced in R20050617 release).
|
|
|
|
|
Added the function unblkdiag
to automatically detect and convert block diagonal
semidefiniteness constraints.
|
|
Added the function dissect
to automatically transform large sparse structured
semidefiniteness constraints to smaller block diagonal
constraints by introducing additional variables.
|
|
The experimental cutting plane solver for mixed integer SDP is now
somewhat official.
Just use the solver-tag cutsdp . Note, this module
requires a fast MILP solver, and is only useful for very
special problem classes.
|
|
Added the command blkvar to simplify
handling of block matrices.
|
|
Multiple sdpvar objects (placed
in a cell) can now be created in one call to sdpvar
by using vector valued dimension.
|
|
|
|
Additional option for removing inequalities, see
sdpsettings.
|
|
|
|
Updated default values of DSDP
options and fixed detection of feasibility/unboundness reports.
|
|
Fixed a bug in YALMIPs mixed integer solver, causing crashes when applied to
higher order nonlinear problems.
|
|
|
|
Substantial performance improvements in manipulating problems with
a large number of SDP cones.
|
|
|
Fixed bug in integer programming module
(caused problems when mixed integer quadratic programs where
solved using a semidefinite programming solver)
|
|
|
Fixed bug in sum of squares module (may have caused wrong results in some obscure large-scale
cases.)
|
|
|
|
Fixed bug in integer programming module
(could cause sub-optimal solution for binary QP problems being
reported as optimal.)
|
|
|
|
|
Added post-processing steps to the sum of squares module. Improves quality of decompositions substantially
in some cases.
|
|
Dualize can now be applied to problems with second order cone constraints.
|
|
Efficient handling of translated primal cones in dualize.
|
|
YALMIP can now automatically update values of primal cones and free variables
when a dualized problem has been solved.
|
|
Added a module for solving geometric
programs using the solver
fmincon (the problem is convexified using the
standard logarithmic variable transformation). For this
solver to be invoked, use the solver tag 'fmincon-geometric' .
This feature is mainly intended for the
fmincon solver in the MathWorks Optimization Toolbox.
It may work with the
fmincon solver in
TOMLAB, but this
has not been tested to any larger extent.
|
|
The built-in integer branch-and-bound solver 'bnb' can now be
applied to arbitrary polynomial nonlinear problems.
Note, there are no global guarantees if the relaxed continuous
problem is nonconvex. Branching is only performed w.r.t
integrality.
|
|
The built-in bilinear branch-and-bound solver 'bmibnb' now supports nonlinear objective functions.
|
|
The built-in bilinear branch-and-bound solver 'bmibnb' now supports discrete variables.
|
|
Logic constraints are supported (initial experimental code). Check out the
logic examples.
|
|
Problems defined with kyp objects
now (most often) exploit structured matrix data when
LMILAB
is used.
|
|
Added support for the QP solver
MEXCLP.
|
|
Added support for the MILP solver LP_SOLVE.
|
|
Added support for the QP solver BPMPD.
|
|
Added support for the rank-constraint SDP solver LMIRANK.
|
|
|
|
The option 'sos.clean' is now eps by default (used to be 1e-4). This may lead to more spurious terms in sum of squares decompositions, but improves robustness w.r.t to
different SDP solvers.
|
|
Fixed
geomean2 for vectors
(enables maximization of products of positive variables).
|
|
Extended
geomean2 to support
complex data.
|
|
More efficient use of given feasible initial solution in the global
solver.
|
|
Fixed bug that caused sign-errors on duals computed using
MOSEK.
|
|
Fixed bug for complex-valued second order cone
problems in SeDuMi.
|
|
|
|
Various fixed related to systems with
TOMLAB installed.
|
|
|
|
General performance improvements (massive improvements for large-scale problems and nonlinear expressions)
|
|
|
Initial implementation of min/ max/ abs/ norm/geomean2/sumk/sumabsk. YALMIP will perform
an (conservative) analysis of the optimization problem to ensure
that these operators are used in a way that guarantee convexity,
and automatically generate a convex representation if possible. Check the
example for typical usage, limitations and pit-falls.
Additional (epigraph or hypograph representable) nonlinear operators can easily be added by the user.
|
|
|
|
Extended the moment relaxation module,
including polynomial
equalities, polynomial semidefinite inequalities, extraction of global
solutions and better integration in YALMIP.
|
|
Improved sum of squares module (better support for nonlinearly
parameterized programs, improved kernel and image representation
models).
|
|
Added the function dualize to convert a problem given in primal SDP form (kernel representation) to a dual form (which is the standard model in YALMIP).
Some simple examples are available.
A related (less useful) function
primalize is also available.
|
|
Polynomial programs are automatically reduced to bilinear programs when
PENBMI
or the global solver is used.
|
|
|
|
Updated for
PENBMI
2 (if you still use version 1, make sure to read the
FAQ).
|
|
Improved memory use for huge SDP problems (matrix dimensions in the order of thousands).
|
|
Fixed bug that caused MAXDET problems to become
unnecessarily
large when solved using standard SDP solver (all MAXDET problems where treated as complex-valued).
|
|
Fixed some algorithmic bugs in the built-in integer solver (caused
severe performance degradation).
|
|
Changed default settings for
the
CSDP
options.
|
|
|
|
Fix in the
MEXPRESS interface (caused severe numerical problems in recent versions of
XPRESS).
|
|
|
When a solver not supports equality constraints, the constraints are
now by default always eliminated using double-sided inequalities (this
was a bit inconsistent before, YALMIP used a reduced basis for some
solvers). Note that some solvers behave much better with a reduced
basis (see sdpsettings for
details).
|
|
Fixed small bug in the sum of squares module
(caused a crash in some rare cases).
|
|
Fixed issues with the option 'savedebug' .
|
|
Fixed issues with saving and loading objects.
|
|
|
|
|
|
|
|
|
|
|
|
|
Added support for the SDP solver
SDPLR.
|
|
Added function saveampl
to export (very) simple
AMPL models.
|
|
Added function export
to export YALMIP models to various solver formats.
|
|
Added function cut
to add user-specified cuts in the global BMI solver.
|
|
Global BMI problems can be solved (hrm...yea right...) without a local
BMI solver.
|
|
Added simple command-line syntax for defining multiple scalar
sdpvar variables.
|
|
Better support for sum-of-squares decompositions in
checkset.
|
|
Added function lowrank
to declare low-rank data in semidefinite constraints.
|
|
|
|
Numerous fixes in the interface to KYPD
(may have caused erroneous solutions).
|
|
Changed the syntax slightly in
solvemoment to make it more consistent with other commands (may
brake old code).
|
|
|
|
Added some missed options for
SeDuMi and updated default values to those recommended in version
1.05R5.
|
|
Minor updates for MATLAB 7
|
|
|
|
Fixed a bug which in rare cases caused YALMIP to categorize standard
continuous problems as a mixed integer problem.
|
|
Fixed some more issues on indexing of
set and sdpvar objects.
|
|
|
|
Fixed problems in loading saved sdpvar
and set objects.
|
|
|
|
Reduced memory usage in branch and bound codes.
|
|
Fixed some small bugs in the mixed integer programming (caused problems
for mixed integer geometric programs).
|
|
Substantial performance improvements in manipulation of
sdpvar variables.
|
|
Fixed several bugs in the interface to the SOCP solver in
MOSEK.
|
|
|
Determinant maximization problems can now be solved using any SDP
solver.
|
|
Global bilinear programming (alpha version, slow and unstable...).
|
|
Extended mixed integer code to include posynomial geometric programming
(requires latest MOSEK patches, not
seriously tested).
|
|
|
|
Added interface to the binary LP-solver
bintprog.
|
|
sdpvar can be called with only
one argument (simplifies definition of symmetric matrices and scalars).
|
|
Determinant ( det ) overloaded on
sdpvar objects.
|
|
Fixed bug in internal mixed integer solver (may have caused YALMIP
to miss the optimal solution).
|
|
Fixed bug in interface to PENBMI
(may have caused erroneous solutions in some rare cases).
|
|
Fixed bug on complex-valued multiplication (may have caused erroneous
solutions).
|
|
Fixed bug on dual variables (may have reported wrong values in some
rare cases).
|
|
Fixed bug on dual variables in SDPT3
(gave completely wrong answers).
|
|
New, more consistent, syntax in solvesos.
|
|
Fixed bugs in loading and saving of set
and sdpvar objects..
|
|
Fixed bug on overloaded >= operator.
|
|
Fixed various issues on indexing of set
and sdpvar objects.
|
|
Fixed bug on purely binary problems using
MOSEK.
|
|
Fixed various bugs to enable nonlinear objective functions in moment
relaxations.
|
|
Fixed bug in interface to SDPT3 on
UNIX.
|
|
Updated for forthcoming SDPT3 3.1.
|
|
|
|
Various fixes for MATLAB 5.3.
|
|
YALMIP now works with LMILAB in MATLAB 5.3.
|
|
YALMIP does not mean Yet Another LMI Parser any more.
|
|
|
Polynomial expressions.
|
|
Quadratic objective functions.
|
|
|
|
Added possibility to define KYP-related constraints ( kyp)
and related interface to the specialized code
KYPD.
|
|
Multiparametric linear and quadratic programming using
MPT-Toolbox.
|
|
Integer and binary variables.
|
|
Posynomial geometric programming.
|
|
An internal (wickedly slow) branch-and bound solver for mixed integer
conic programs (MILP, MIQP, MISOCP, MISDP).
|
|
Sum-of-squares decompositions.
|
|
Moment-based polynomial programming.
|
|
|
|
Plot the feasible set of a constraint.
|
|
|
|
Imroved support for subsref in set
(previously called lmi) and
sdpvar objects.
|
|
New feature in the option removeequalities in
sdpsettings.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|