# myTOMLAB  
*From the Author
*Contact us

News Logo

October 18, 2022

Version 8.9 released.

Tomlab /KNITRO:
  • New version 13.1
Tomlab /MAD:
  • Better support for singleton expansion

March 18, 2022

Version 8.8 released.

Tomlab /CPLEX:
  • New version 20.1
Tomlab /KNITRO:
  • New version 13.0
Tomlab /SNOPT:
  • Fix rare/random crash on recent Matlab versions

Sep 17th, 2020

Version 8.7 released.

Tomlab /CPLEX:
  • New version 12.10
Tomlab /KNITRO:
  • New version 12.2
Tomlab /NLPQL and /MISQP:
  • Versions 5.03 and 7.31 respectively
  • Improve log to screen

Nov 15th, 2019

Version 8.6 released.

Tomlab /BASE:
  • Open-source derivative-free solvers BFO and SID-PSM added
  • Fix bug causing dense all-zero matrix allocation (tomSym)
  • Singleton expansion improved (tomSym)
Tomlab /MAD:
  • Singleton expansion improved
Tomlab /QMINOS:
  • Quad-precision MINOS introduced
Tomlab /CPLEX:
  • New version 12.9
Tomlab /KNITRO:
  • New version 12

Sep 1st 2018

Version 8.5 released.

  • Re-write feature in tomSym for handling objectives with abs-like constructs
  • Fixed inaccurate computation of Chebyshev node points.
  • Version 7.7 (SNOPT and SQOPT)
  • New solver package for dense nonlinear optimization
  • New version 12.8
  • New solver version 8
TOMLAB /Knitro:
  • New version 11.0

June 30th 2016

Version 8.2 released.

  • Monitor callback feature added (progress/stop optimization feature)
  • New version based on Artelys Knitro 10.1
  • Spline-based collocation added as an alternative to pseudospectral collocation.
  • Updated to new version with improved performance
  • New solver package: Mixed Integer Distributed Ant Colony Optimization
  • Improved bessel function compatibility, add interp2 for newer Matlab versions
  • New feature: Detection and differentiation of strict integer variables (not continuously differentiable)

Jul 4th 2014

Version 8.0 released.

  • glcCluster: New faster cluster algorithm implemented, several bug fixes and speed improvment when used as sub solver with the CGO TB.
  • multiMin: Several bug fixes and minor improvements. Two major loops may run in parallel on all threads available using parfor.
  • One con_prob example fixed for parfor compatibility.
  • optim/linprog: fixed problem when called from GA solver.
  • lpSimplex: scaling improved.

  • Version 12.5 integrated.
  • Avoid crash for problems with exclusively quadratic constraints.
  • Number of iterations is always properly returned.

  • Based on Gurobi 5.6.3 (except Linux 32-bit, which is 5.0.2).


Complete revision. Nine different ways of function transformation now possible. All solvers now prepared for use with multithreading and parfor usage. Optionally, all solvers will try to find all local and global optima. Also possible to input known minima to avoid sampling close to these. All solvers use a new point selection algorithm to avoid ill-conditioning. The CGO sub solvers glcCluster and multiMin now solves sub problems faster. All CGO solvers now use a random resampling strategy if the major algorithm cannot find a new well-separated point.
  • rbfSolve: Improved speed. Add as trial points an optional number of additional global/local minima obtained during RBF surface and/or target value optimization.
  • ego: Now 4 different EGO type of algorithms implemented, with 6 types of sample criteria functions. Option to use cross validation to find best possible function transformation. Option to speed up DACE parameter fitting. Add as trial points an optional number of additional global/local minima obtained during surface and/or sample criteria optimization.
  • arbfmip: Revised to run much faster on higher dimensional problems. Add as trial points an optional number of additional global/local minima obtained during RBF surface and/or target value optimization.

  • Based on KNITRO 9.0.1

  • New solver package for Dense Mixed-Integer Nonlinear Programming (MINLP). Includes the solver MISQP and sub solvers MIQL, NLPQL and QL, developed by Klaus Schittkowski.


New solver package for Mixed-Integer Nonlinear Optimization (MINLP) and Global (non-convex) Optimization (GO).
  • Solver ecpMINLP implements an Extended Cutting Plane algorithm for convex and pseudo-convex MINLP by Westerlund et al.
  • Solver stoaMINLP implements a Single-search Tree Outer Approximation algorithm for both convex and nonconvex MINLP based on research by Leyffer et al.
  • Solver multiMin is a Multistart GO algorithm with simple mixed-integer handling developed by Holmstrom.
  • Solver multiMINLP implements an algorithm by Holmstrom for small-scale non-convex MINLP where the integer combinations are huge, but relaxation of the integer constraints is possible.

Aug 23rd 2012

Version 7.9 released.

  • Trial licenses work better with parallel setups (parfor etc)
  • Fixed bug in Optimization Toolbox interfaces related to parameters

  • Fix wrong gradient returned after solve

  • Responsiveness to Ctrl-C improved

  • Improvements in speed and memory efficency
  • Support for more functions with tomArray

  • Simplified multi-phase notation. See jumboContainerCrane example.

Dec 16th 2011

Version 7.8 released.

  • Version 7.7 released.
  • Improvement for MPEC problems solved with numerically estimated constraint gradients.

  • Bug fix in presolve feature, improved stability for continuous problems.

  • Now based on GUROBI 4.6.

  • Based on KNITRO 8.0.

Jun 7th 2011

Version 7.7 released.

  • Version 7.7 released.

  • CPLEX 12.2 now included.
  • Kappa statistics available for MIP problems solved with simplex node algorithms.

  • Now based on GUROBI 4.5.

  • Based on KNITRO 7.0.
  • Fixed crashes that occured on scalar problems under certain conditions.

  • Update for maximization problems.

Nov 24th 2010

Version 7.6 released.

  • bintprog updated to handle more solvers.
  • tomSym updated for complimentary optimization problems.
  • milpSolve based on lp_solve
  • Several optimizations and speedups in tomSym and PROPT.

  • Supports CTRL-C for interrupting MIQP optimizations.

  • Now based on GUROBI 4.0.
  • New version supports QP, MIQP in addition to LP, MIP.
  • Improved performance for all problem types.
  • Bugs removed related to log files.
  • Return codes updated.

Oct 1st 2010

Version 7.5 released.

  • multiMin now supports problem with infinite bounds.
  • mipSolve no longer generating save files.
  • log2 added to tomSym.
  • glcCluster updated to avoid harddrive access.

  • Improve solution pool feature for problems that are reduced to nothing by presolve.

  • An issue with transpose has been resolved.

  • Two new test cases :

        Metabolic Pathways.
        Temperature Control at Night.
  • Optimal control problems with binary or integer states, controls or constants is now possible.

  • Version 3.0 released. Changes from version 2.0 include:
  • Parallel Barrier LP method added, with corresponding callback and crossover algorithm.
  • MIP performance increased. Symmetry handling, improved cutting planes, additional MIP heuristics.
  • MIP memory management improved.
  • Possible to retrieve any feasible solution found during branch-and-cut search.
  • MipNode callback added.
  • Additional presolve reductions improve LP and MIP performance.
  • Several parameters and model attributes added.

  • Bug related to problems with a special form of nonlinear constraints removed.
  • Memory management improvements.

  • Updated for large-scale cases.

Mar 26th 2010

Version 7.4 released.

  • interp2 and interpn are now available for tomSym.
  • minlpSolve and mipSolve:

        Change logic to always compute the duality gap when an integer solution found. Now input tolerance DualGap = 1 always gives a stop after the first integer solution found.
        The duality gap is printed in the exit text. Correction of sign in duality gap computation.

  • glcCluster:

        Corrected one of the tests that avoids the use of the same initial point twice.

  • EGO:

        Major revision of the convergence test based on the scaled expected improvement.
        Now EGO always picks the point with the lowest function value if two points are very close.
        Screen printing updates.

  • xIP checked in more detail.
  • Unexpected solution pool related crash avoided for some special cases.
  • Screen printing properly disabled when running several threads.

  • Several new test cases:

        Orbit Raising Problem - Minimum Time.
        Orbit Raising Problem - Maximum Radius.
        Maximum Crossrange Space Shuttle Reentry Problem.

  • New scaling module enables the solution to a wider range of problems.

  • Backend library minor update to 2.0.2.

  • Enabled options BLASOPTION and BLASOPTIONLIB, facilitating using external BLAS library if desired.

Dec 7th 2009

Version 7.3 released.

  • Improved derivatives for expressions involving large matrices in tomSym. Some second derivatives that previously resulted in too large sparse matrices may now be computed.
  • gammainc now supported for tomSym.
  • Help review for assign routines.
  • A much more stable and robust minlpSolve, that now is usable for nonconvex and convex MINLPs with rather many IP combinations:

        Complete solver revision.
        Solves MINLP with either nonconvex or convex (set flag Convex) NLP subproblems.
        May handle a large number of IP combinations by branch and bound, thereby able to solve e.g. Trim Loss problems.
        Now possible to warm start the solver, using field Prob.minlpSolve, and calling WarmStartGLOBAL.
        Saves a mat file every iteration. This enables warm starts.
        RandState is saved and input to multiMin, giving identical runs in case of Warm Start of a nonconvex problem.
        It is possible to give an overestimate fIP. All nodes with f > fIP will be ignored during the search.
        Improved node selection strategies implemented as NodeSel 0-2.
        Old strategies still available as NodeSel 3-5.
        Added rounding heuristic and knapsack heuristic, new input.
        Prob.ROUNDH (default 0) and Prob.KNAPSACK (default 0).
        If using the heuristics, probably NodeSel = 1 better than 2.

  • multiMINLP a new TOMLAB solver for nonconvex INLP (or MINLP) where the number of IP combinations are too many for glcDirect to handle:

        Aimed for small size (dim < 50) black-box MINLP or MINLP with 1st derivatives available.
        Targets problems where the number of integer combinations are huge (>> 10000).
        Solves MINLP problem in small subareas, determined by solving relaxed NLP problem with multiMin.
        Mainly used as subsolver for the CGO solvers.

  • Several improvements for multiMin, especially for MINLP, but also global nonconvex problems. Now returns all global minima found. Appears very robust, and with use of the new warm start very seldom fails:

        Complete revision, now efficient for nonconvex MINLP with not too many integer combinations (order of hundreds).
        Now classifies and sorts every solution found, removes duplicate solutions.
        When doing warm start, multiMin determines xInit number of initial x_0, all not too close to previous x_0 and x_k.
        multiMin now handles MINLP with number of IP combinations of the order of hundreds.

  • mipSolve greatly improved:

        Complete solver revision.
        Input field Prob.Solver.Alg is changed to Prob.MIP.NodeSel (similar field as minlpSolve).
        Improved node selection strategies implemented as NodeSel 0-2.
        Old strategies still available as NodeSel 3-5.
        mipSolve now possible to warm start, using field Prob.mipSolve and calling WarmStartGLOBAL.
        mipSolve also saves a mat file mipSolveSave.mat every SaveFreq iteration.
        Input field Prob.MIP.SaveFreq gives the frequency (default 100). Thereby enabling warm starts.
        Now possible to give an overestimate fIP. All nodes with f > fIP will be ignored.
        Added rounding heuristic, improved knapsack heuristic, new input.
        Prob.ROUNDH (default 1) and Prob.KNAPSACK (default 1).

  • Now uses one internal routine to compute alpha in cycle, and one external routine getfnStar to compute fnStar in cycle.
  • Fixed crash for Prob.CGO.idea = 2. Must not use new cgolib, because gn_g is not available for the idea=2 (alpha type) subproblem.

  • Better error messages.
  • Isoperimetric constraint problem added.
  • Two-phase penicillin production (fed-batch fermentor) added.

  • Memory problem for GENO corrected.

  • New version 2.0
  • Greatly improved simplex and MIP performance.
  • Possibility to use node files for MIP problems.
  • New parameters and cuts available.

Aug 18th 2009

Version 7.2 released.

  • Support for more functions in tomSym, including: atan2, gamma, gammaln, psi, erf, erfc, erfinv, erfcinv, diag, spdiags, sparse, cumsum, eps, floor, round, fix, mod, dblquad.
  • Support for subscripted assignment to zero elements of tomSym objects.
  • Support for mixed complementary problems (MCP) in tomSym, when used with KNITRO.
  • More than 80 tomSym test cases are now included.
  • Significant speed-ups for LP problems with tomSym.
  • Formal support for multi-index optimization added. See for example:


  • Improved error messages for tomSym.
  • Major revision of TOMLAB manual for the tomSym sections (Section 4.3 and Appendix C).
  • tomSym now supports least squares (L2-norm), least absolute deviations (L1-norm), and mid-range (L-infinity norm) problems.
  • Array concatenation corrected for tomSym.
  • Many general example updates to avoid variable and function conflicts.
  • Several speed-ups for linear least squares and quadratic programming problems.
  • tomSym problem with more than 32 nested parentheses removed.
  • Use of automatic differentiation with tomSym now assumes nonlinear programming problem.
  • Problem diagnostic for tomSym safeguarded for memory limitations.
  • Possible to avoid generatation of gradient and Jacobian for tomSym, in case a solver such as LGO or glcDirect is used.

  • New MILP solver with unlimited core/cpu usage at no additional cost.

  • besselj, bessely and besselk added.
  • Screen print out removed.

  • Possible to use CGO with function handles and anonymous/nested functions.

  • Version 12.1 released.
  • Possible to load LP and SAV files.
  • The standard license now includes unlimited shared memory core/cpu usage.

  • New version of SNOPT (7.2-9) now embedded in TOMLAB.

  • Minor example updates.
  • Help update for tomControls.
  • Minimum Energy Orbit Transfer example added.

  • Updates to the return codes. Several manual corrections and improvements.

Aug 6th 2009

Multi-Platform Installations Supported.

Tomlab Optimization has switched to BITROCK to enable support for standardized installation procedures for Windows (32/64-bit), Linux (32/64-bit) and Mac OS X.

Mar 25th 2009

Version 7.1 released.

  • Many general tomSym improvements.
  • tomSym help updated for sym2prob - now an options structure input.
  • tomSym Screen print-out removed when problem type pre-selected.
  • tomSym equation parser updated to handle redundant constraints (zero constraints).
  • triu and tril added for tomSym.
  • binbin2lin and bincont2lin examples added to the TOMLAB quickguide.
  • Error messages improved for license problems.
  • Return codes corrected for linprog and quadprog.
  • Numerical differentiation screen printing removed.
  • Problem when solving MIQQ problems of varying size removed.
  • Now compatible with MATLAB R2009a.

  • Updates for arbfMIP and other solvers.

  • Error stack exhaust fixed.
  • Examples for using CPLEX solver tuning methods and solution pool features added.
  • Return code issues corrected for non-convex QP problems.
  • Return relative MIP gap in cpxRetVec (global).
  • Several manual updates related to global parameters.
  • Size problem with solution vector corrected.
  • Duplicate solutions in pool corrected.
  • Thread default corrected.
  • Fixed issue when giving too many xIP values.
  • Detects all-zero quadratic constraints.
  • Using solution pool could return too few solutions.

  • snoptA reviewed.
  • Bug in SQOPT causing a crash removed.

  • KNITRO 6.0 now embedded in TOMLAB.
  • New MINLP support features and other updates.

  • Removed bug which would sometimes yield incorrect size matrix when expanding scalar multiplied by matrix using collocate() or atPoints().
  • Speed improvements for collocate() on expressions involving sum(), norm(), cross(), and certain other functions.
  • Less cluttered rendering of tomStates and tomControls when symbolic expressions are displayed.
  • mcollocate extended to includes the initial and final points.
  • Bug in tomPhase removed, allowing for more custom user inputs.
  • More examples added (supplied by Gerard Van Willigenburg, Willem De Koning and Russ Tedrake) - now a total of 106!

        Acrobot example
        Five-link MK2 robot
        Food sterilization
        Green house control system
        Jumbo container crane
        Maximum orbit transfer
        Optimal Parametric Sensitivity
        Radio telescope positioning
        Temperature control
        Third order system
        Two-Link robot
        Zermelos problem (version 1)
        Zermelos problem (version 2)

  • logical variables (constants) now handled.

Nov 17th 2008

TOMLAB 7 finally out!
The new platform features among other things:
  • TomSym now included!
  • PrintResult report corrected when using numerical derivatives for Jacobian.
  • multiMin sub-solver support for non-convex QP problems expanded.
  • addPwLinFun updated.
  • Quickguide example added for piece-wise linear programming (two versions).
  • bintprog updated to work more smoothly.
  • Many updates to Linux versions for dependencies.
  • Now possible to use LINPROG as subsolver for multi-layer optimization problems.
  • Return codes from LINPROG updated.
  • Unnecessary calls to checkType, LineParamDef and more removed.

  • The following functions added: DBLQUAD
  • Updates to ZEROS and ONES to handle scalar case.
  • SUM screen print out removed.

  • Four new radial basis interpolation functions added to rbfSolve and arbfMIP. Besides cubic and thin plate spline, the solvers may use linear, multiquadric, inverse multiquadric and Gaussian RBFs.
  • Warm start information is saved every iteration, which makes it possible for the user to break the run, and continue with a warm start with any CGO solver with new parameter settings.
  • EGO is using five new sample criterion functions, beside expected improvement: Kushner's criterion, Lower confidence bounding, Generalized expected improvement, Maximum variance and the Watson and Barnes criteria 2.
  • Many updates for integer programming.
  • Memory allocation problem fixed.
  • Many documentation updates.

  • Version 11.2 embedded.
  • Problem when using multiple threads corrected.
  • TUNINGTILIM data type problem fixed.
  • Tuning tool updated to handle TUNE = 2 correctly.
  • Solution pool capabilities fully activated.

  • snoptA released with examples.
  • SNOPT/SQOPT 6 are now obsolete and completely replaced by SNOPT/SQOPT 7.

  • First derivative checking feature added.
  • M-file callback feature fixed.


Jun 10th 2008

TOMLAB version 6.1 released.
  • glcDirect printing updated.
  • glcCluster using less memory for problems with integer variables.
  • mipSolve using subsolver in the following order (determined by license): MINOS, BQPD, XA, MILPSOLVE.
  • Substantial updates to Linux distributions (library dependencies and compilers).
  • glcCluster updated to use maxLocalTry.
  • Problems when running quadratically constrained problems of varying sizes fixed.
  • nlpSolve updated to handle LP subproblems.
  • NumDiff and ConsDiff automtically set to correct values when empty.
  • Second order information not computed for problems without analytical derivatives.
  • nlpSolve now has 500 iterations as default.
  • Optimization toolbox interface updated.
  • Now poossible to use MAD with multiMin.
  • linRatSolve automatically purging binary and integer variables.
  • miqqAssign updated to support problems without quadratic constraints.
  • bincont2lin now taking lower bounds into account.
  • NumDiff and ConsDiff automtically set to correct values when empty.
  • Change to relative constraint violation instead of absolute constraint violation.
  • Conflict between relative constraint violation in DIRECT methods and absolute constraint violation in local search using consviolation solved. Now consviolation does relative constraint violation computations as well.
  • glcCluster trying to utilize derivatives gave wrong results if given routines were dummy routines (e.g. for CGO subsolvers).
  • WarmDEFGLOBAL updated with CGO solvers arbfMIP and rbfSolve.
  • Major speed-ups for mipSolve, conSolve and other Base Module solvers.

  • MAC OS X (Intel) now supported.
  • Advanced presolve functionality enabled.
  • Help updated.
  • Problem with incorrect objective reported fixed.
  • b_L and b_U verified to be dense.
  • Change obsolete parameter MIPSTART into ADVIND.
  • Screen printing disabled when using more than 1 thread.

  • The following functions added: FFT, FFTN, IFFT, FFTSHIFT, IFFTSHIFT.
  • New website for MAD:
  • TRANSPOSE updated for improved speed.
  • IMAG and REAL added.
  • General updates to derivvec class.
  • MIN updated.

  • Version 5.2 embedded.
  • New options for multi-start feature: MS_NUMTOSAVE, MS_SAVETOL and MS_TERMINATE.
  • New options for file printing: OUTDIR and OUTAPPEND.

  • PENSDP and PENBMI added for Linux 64-bit.

  • Defaults updated for LPOPT, QPOPT, SNOPT7, SNOPT, SQOPT, SQOPT7.

  • Fixed bug when x_0 present.

  • Writing solution file regardless of ExitFlag.

  • Released (Lead Developer is Per Rutquist - Matlab Programming Contest champion).
  • A complete framework for optimal control (ODE and DAE) and parameter estimation.
  • About 100 fully functional test cases.
  • See the dedicated webpage here:
  • 6 years of optimal control experience has proven the software unbeatable in comparison to other modules.

  • Use relative constraint violation instead of absolute.
  • Total revision of experimental design methods (now 11 design methods). Updated values for Percent and nSample and new input parameters nTrial and CLHMethod.
  • Help updated for all solvers.
  • EGO:

        Complete revision and various bug fixes.
        New options for DACE surface estimation.

  • rbfSolve:

        New revision and many bug fixes.
        Revised for pure integer problems. Minimum of surface for pure IP often gives the same minimal point. A new special algorithm finds a neighboring integer point if this occurs.
        New Inform codes added.
        Speed up gateway routines for sub problems.

  • arbfMIP:

        First release of code implementing the new Adaptive Radial Basis Algorithm (ARBF).
        The algorithm has been discussed by the TOMLAB team in recent publications in the Journal of Global Optimization and Optimization and Engineering.

Jan 31st 2008

TOMLAB® is now a registered trademark of Tomlab Optimization. The registration can be found at the United States Patent and Trademark Office website.

Dec 28th 2007

TOMLAB version 6.0 released.
  • All short-circuit and/or statements reversed to standard | and &.
  • It is now possible to define the problem with the following functions:

        Nested function
        Anonymous function
        Subfunctions defined within main file
        Regular m-files referenced by function handle or string

  • glcDirect error reporting updated when using function handles.
  • Tfmin updated and added to manual.
  • Several Base Module solvers updated to handle non-symmetric matrices.
  • glcCluster warm start updated.
  • Tfzero help updated (not possible to use gradient) and added to manual.
  • glcCluster now using Prob.x_0 and Prob.X0 as initial starting points for local searches.
  • multiMin - new default. Now using min(3000,30*n) local searches.
  • ucDemo updated to TOMLAB format.
  • tomGui and other GUIs removed from distribution.
  • Major updates to the general TOMLAB manual.
  • Manual updates for MAD, CPLEX, KNITRO, MINLP, SOL.

  • Updated SOL manuals with regards to basis files.

  • Problem removed when no nonlinear constraints are present.
  • Function evaluation report corrected.

  • Updated to handle LCP, QCP and MCP problems.

  • CPLEX 11 now embedded in the official release.
  • New tuning tool that helps the user minimize the execution time in production.
  • Parallel mode possible for faster solution of mixed-integer models.
  • Possible to collect many solutions in a pool.
  • A wider variety of nonconvex quadratic constraints are now automatically handled.
  • No longer printing error message when empty linear constraint matrix.
  • CPLEX no longer crashing when providing xIP for MILP/MIQP problems.
  • Using embedded license by default.
  • Branching priority added to input variables.
  • Possible to give branching directions for individual variables.

  • Updated interp routines to work for most future Matlab releases.
  • AD for second order constraints working.

  • Now works with Matlab 7.5.

  • Available for Windows 64-bit.
  • Printing removed when empty linear constraints are included.

  • LSGRG2 printing updated.

  • Problem fixed for R2007a and b.

  • Bug when using glcCluster as subsolver removed.
  • rbfSolve and arbfmip (not released yet) updated for problems with stationary sample points.
  • Three new experimental designs (-997, -998, -999 (revised)). Strategy -999 is no longer called the Gutmann strategy.
  • Updated when using multiMin as global solver.
  • New REPLACE option for rbfSolve and arbfmip.
  • Max Lipschitz constant LipMx estimated using the initial set of points X and updated in every iteration.
  • New defaults for box-constrained problems (REPLACE = 5, SCALE = 0).
  • AddSurfMin new parameter for checking additional interior local minima.
  • New parameter when using multiMin.
  • RandState modified for EGO when no warm start.

Sep 7th 2007

TOMLAB version 5.9 released.
  • Bug in PrintAssign fixed.
  • Several updates to lsqcurvefit.
  • conSolve updated gradient handling.
  • Compatible with Matlab 7.5 (R2007b).
  • clsSolve updated for Broydens method - Jacobian set correctly initially.
  • Bug in qpSolve removed.
  • ucSolve printing updated.
  • glcDirect updated to take 1 or 2 inputs.
  • multiMin updated to enable warm start of solution process.

  • SNOPT 7.2-5 (May 2007) available.

  • A patch for TOMLAB /PENSDP and TOMLAB /PENBMI that works with Matlab 7.5 is available here.

  • Now much easier to formulate problem and evaluate solution.
  • Solver speed performance greatly improved.
  • Bug causing memory problems fixed.

  • Memory bug fixed.
  • Improved Hessian handling for large problems.

  • New tool for adding piece-wise linear functions and constraints to an existing model.
  • User cuts can now be used in non-parallel mode.
  • Incumbent callback available.

  • Help updated.

Jun 28th 2007

TOMLAB version 5.8 released.
  • milpsolve MaxCPU parameter added.
  • Updates implemented for MAC OS X (Intel).
  • Warm start code updated for most solvers.
  • License not found print-out from MAD removed.
  • Possible to generate a print file from TLSQR.
  • estBestHessian updated for use with MAD.
  • License problems for Windows 2000 and Windows Vista corrected.
  • Full system using short-circuit and and or (no longer compatible with Matlab 6.0).
  • Safer memory handling in tomsol.

  • Problem with objective for linear problems resolved.
  • Conflict refiner can now generate a print file.
  • Callback added: cpxcb_INCUMBENT. Possible to store all integer solutions in Matlab.
  • Review of all return and error codes.

  • Minor incompatibility problem fixed with other toolboxes.

  • NPSOL return code and text modified.
  • LPOPT and QPOPT updated for warm start.
  • NPSOL updated for automatic differentiation of nonlinear constraints.

  • Updates to manual.
  • Updated to handle empty constraint Jacobian.

  • Notation updated to match standard.
  • Updated to handle empty constraint Jacobian.

  • Memory problem corrected.

  • Bug in Matlab prevents object oriented code from working properly (avoid Matlab 7.4 with MAD)
  • Added function: tanh, sinh, cosh, coth, atanh, acoth and asinh.
  • userfunc.m added as template for user functions.
  • interp1 and interp2 compatible with Matlab 7.4.

Mar 7th 2007

TOMLAB version 5.7 released.
  • Intel based MAC supported.
  • startup modified for backward compatibility.
  • Better error messages added to startup.
  • ADMAT no longer supported.
  • Assign routines standardized.
  • QP Block optimization problem added to quickguide.
  • checkDerivs updated to handle HessPattern correctly when given.
  • slsSolve updated to always use SNOPT by default.
  • Matlab 7.4 supported - a separate installation is needed.
  • bincont2lin updated and corrected.
  • Updates to Tfzero, glbFast and glcFast.
  • binbin2lin updated.
  • Complete system review for handling of integer variables.
  • Manual updated for most product suites.

  • Inputs no longer modified when calling cplex directly.
  • Added checks for NaN and Inf inputs.
  • Safe-guard added for IntVars.
  • Bug fixed when quadratic objective is empty (for problems with quadratic constraints).

  • Memory overflow corrected for SQOPT6 and SQOPT7.
  • Manual updates for all solvers.
  • LU basis factor storage memory bug fixed for MINOS.
  • moremem parameter added for MINOS solver. Updated for SNOPT and SQOPT.

  • Iteration limit added to minlpBB.

  • New version 5.1.1 released.
  • Windows 64-bit supported as platform.
  • MPEC examples updated.
  • Shared error added for missing DLL paths.

  • min updated.

  • New option RANDOM_NUMBER_SEED available.

Jan 23rd 2007

Dr. Naz Bedrossian (a TOMLAB user since 2002) with Draper Laboratory in Houston, TX has reported that TOMLAB results were successfully used at the international space station. For more information please visit this link. Further details about the use of TOMLAB /SNOPT will be published in the future.

Dec 12th 2006

The first TOMLAB-managed currency trading fund in Sweden is now open for investments. For more information visit eTurn (Swedish only).

Dec 5th 2006

TOMLAB version 5.6 released.
  • Separate 64-bit versions developed for Matlab 7.3 (R2006b) and later.
  • Several updates the TOMLAB test suite.
  • Writing to MAT-files from global solvers (glb/glcDirect, etc) safer, avoiding losing results after long runs.
  • Minor help updates to m-files.
  • qpSolve updated to work with nonlinear subsolvers.

  • interp1, interp2 and max updated.
  • not added to distribution.
  • cat added - possible to do horizontal and vertical concatenation of MAD objects.
  • general bug fixes and updates.

  • Bug related to string buffer fixed.

  • Writing to MAT-files safer, avoiding losing results after long runs.

  • MAXTIME parameter corrected.
  • Updated to handle problems with one dense nonlinear constraint.

  • SNOPT and SNOPT7 no longer making callbacks for problems with pure linear objectives.
  • MINOS updated to handle nonlinear problems with linear objectives.
  • Differences between NPSOL for 32-bit and 64-bit Windows resolved.

Sep 8th 2006

TOMLAB version 5.5 released.
  • Issues with number of inputs for glcDirect and glcCluster resolved.
  • chs_prob updated.
  • Updates to InstallShield installer for Windows (32/64 bit).
  • Several manual updates.
  • Functions for modifying static parts of existing problems added (modify_* and more). See section 4.5 in the main TOMLAB manual.
  • LDO and LDODEMO removed from distribution (available for download from the manuals page).
  • License problem with Matlab R12 resolved.
  • lpSimplex no longer printing by default.
  • TOMLAB version of LSQCURVEFIT bug fixed.
  • clsSolve - Tests added to avoid running problems with nonlinear constraints.

        - Revision of Broydens method (line search, defaults)
        - No longer crash when Jacobian is all zeros.

  • goalSolve - Updates for handling of nonlinear constraints.

        - Handling of Lagrange multipliers updated.

  • Improvements maybe to optimization toolbox plug-in FGOALATTAIN.
  • Tolerances updated for printing.
  • PDSCO now working with problems without linear constraints.
  • Output improved when problems arise.
  • System-wide updates for defaults in assign routines.
  • infSolve and infLinSolve updated for lower bound setting, now possible to get an objective lower than 0.
  • Updates to defaults for line searches used.
  • Consistent handling of convergence to user defined target value.
  • Unnecessary isnan checks removed from all solvers.
  • LINPROG plug-in print outs removed.
  • multiMin adjusted to also handle Matlab 6 and least squares problems.
  • Additional safety added to all assign routines.
  • nlresid and nlfunc no longer calling nlp_c for problems with no nonlinear constraints.

  • A package for static or dynamic, uni- or multi-objective optimization based on a real-coded genetic algorithm.

  • Pre-defined test problems without IntVars can now be run with rbfSolve.
  • Tests if bounds are finite.

  • Manuals updated with return codes for SQOPT.
  • LSSOL/LPSOL/QPOPT/SQOPT updated for problems with no linear constraints.
  • Screen printing buffers corrected for SNOPT7 and SQOPT7.
  • SQOPT now accepting problem with empty linear part.

  • Patterns automatically set for problems with linear or quadratic objective.

  • f_Low now used for the solvers.
  • filterSQP - Lagrange multipliers now calculated correctly.
  • Help updated regarding dense and sparse solver versions.

  • Several general updates and improvements to the system.
  • interp1 and interp2 added to functionality.

Aug 14th 2006

TOMLAB version 5.4 released.
  • Help for assign routines updated.
  • ucSolve help updated.
  • New solver multiMin that obtains multiple local minima from either a given or random set of starting points. A local sub-solver is required.
  • New simpler licensing system - substantial speed-ups for demo users.
  • TSP (Travelling salesman problems) integrated in tomlab\testprob.
  • sqr2 removed from distribution.
  • Updated optimization toolbox interfaces.
  • 21 Test problems added from GENO manual.
  • clsSolve now using TLSQR as subsolver for large-scale NLLS problems.
  • PDCO has been improved.
  • mipSolve convergence improvements.

TOMLAB /CPLEX (and more):
  • Screen printing fixed for problems that are very CPU intensive (bug in Matlab prevented prints from being flushed).

  • Updated help.

Jul 1st 2006

TOMLAB version 5.3 released.
  • No lower limit on integer solution set for glcDirect anymore.
  • BuildMPEC added to the distribution. Support with the KNITRO solver now possible.
  • glcSolve output improved.
  • All pattern estimations updated (estConsPattern, HessPattern, d2cPattern, JacPattern).
  • New quick guide with examples for linear, quadratic and nonlinear complementarity problems.
  • Three new assign routines for complementarity problems.

  • Version 5.0 released.
  • New multistart features.
  • Possibility to solve MPEC problems.

  • Algorithmic convergence greatly improved.
  • Printing improved during optimization process.
  • Rescue procedure implemented.
  • Bug fixed that could cause a crash.

  • Version 1.4 released.
  • Complete system review.
  • Contents.m files added to all folders.
  • Facilities for differentiating through black box code.
  • High-level interfaces allowing easy use of MAD with Matlab's ODE solvers and many Optimization Toolbox solvers.
  • Users guide updated with more examples.

May 12th 2006

TOMLAB version 5.2 released.
  • Windows 64-bit supported for most packages.
  • Several manuals are now available in HTML as well as PDF.
  • A new manual covering all the test problems included with TOMLAB released.
  • All assign routines now checking for crossover bounds.
  • FMINUNC interface updated to handle problems with only an objective function.
  • miqqAssign updated to set c_L and c_U correctly.
  • ls_H updated to handle nonlinear least squares problems more efficiently.
  • glcDirect/glcFast - input checking now fixed.
  • glcCluster - major updates to algorithm.
  • infSolve - now possible to include integer variables.

  • Changed glcFast to glcDirect as default, added glcDirect setup.
  • New inputs nSample and eps_sn.
  • Use new function expDesign for initial experimental design.

  • No longer estimating ConsPattern for large scale problems.

  • SNOPT 7 added to warm start functionality.

  • Unsymmetric QP problems handled more smoothly.

Feb 6th 2006

TOMLAB version 5.1 released.
  • System wide manual updates.
  • Routine lls2qp added to distribution. Now possible to convert a problem with an LLS objective to a problem with a QP objective.
  • Now possible to solve mixed-integer linear least squares problem (test problem added to the Quickguide).
  • expSolve (solver for fitting of sums of positive exponentials) separated into expAssign and expSolve.
  • Possible to call slsSolve, L1Solve, infSolve with tomRun and get normal printing.

  • A new version of EGO released - major updates to the entire algorithm structure - the solver has proven more stable with the new c-libraries.
  • DACEProb is no longer a copy of main Prob.
  • Transformation for EGO callback function ego_f added. Can be controlled with EITRANSFORM.

  • NLSSOL no longer doing isfield check in callback.

  • Verify level default set to -1 for all solver. No derivative checks done.
  • Informs values revised for all solvers. New manual and m-files available.

  • Branch strategies for miqpBB updated. Negative variables were causing problems before.

TOMLAB /CPLEX - Version 10 released.
  • A new more advanced conflict refiner has replaced the IIS features.
  • Possible to give logical (indicator) constraints (i.e. a binary variable controls the constraint).
  • The following parameters have been added/removed/modified:

        COVERS - Now possible to do very aggressive cover cuts.
        PRESLVND - One more option added for node presolve.
        SYMMERTY - Several new options to generate these cuts.
        FEASOPTMODE - Added with many 5 possible settings.
        REPEATPRESOLVE - Controls how to re-apply presolve for MIP models.
        MEMORYEMPHASIS - Memory may be conserved using this parameter.
        NUMERICALEMPHASIS - Caution parameter.
        POLISHTIME - Time spent polishing MIP solution.
        EPRELAX - Control for FEASOPTMODE.


Dec 15th 2005

TOMLAB version v5.0 released:
  • estd2cPattern updated (Prob.N was used instead of Prob.mNonLin).
  • portfolioAbsCon updated.
  • Updated error messages for installation problem.
  • binbin2lin added to the distribution - automatically converts problems with binary products. One can then model with an extra variable right away without re-writing the problem.
  • bincont2lin added to the distribution - converts problems with binary integer/continuous products.
  • New solvers added:

        INFLINSOLVE - solves linear and mixed integer linear minimax problems.
        LINRATSOLVE - solver linear programming problem with a ratio for the objective.
        See the TOMLAB manual for more details.

  • Two new problem types added to the TOMLAB Quickguide.
  • New MaxIter, MaxFunc defaults for glbSolve, glbFast, glcSolve and glcFast (MaxIter = max(5000,n*1000), MaxFunc = max(10000,n*2000)).
  • Assign routines updated with more checks on inputs.
  • mipSolve, cutplane and minlpSolve updated.
  • Model library for linear and mixed-integer programming released. Detailed test cases for the following areas:

        air transportation.
        finance and economics.
        ground transportation.
        loading and cutting problems.
        mining and process industries.
        planning problems.
        public services.
        scheduling problems.
        time tabling.

  • A dynamic routine for generation of global test problems added (GLKS).

  • SQOPT7 updated - an incorrect optimal value was reported for some cases.
  • Alpha release of SNOPT/SQOPT 7.1-1.
  • QN (Quasi-Newton CG) QP Subsolver enabled.

  • No longer estimating gradient at end of run.
  • Random strategy updated.
  • Updated behavior when replacing poor values.

  • Automatically sets LS2PTJ depending on user information.
  • Unused variables removed.
  • Possible to set equality tolerance in Prob.CONOPT.eqTol.

  • xState, bState reported correctly.

  • isfinite, isnan, isvector and isinf added to distribution.
  • Dummy files moved to mad/@dummy.
  • tril, triu, cumsum added to supported functions.
  • reshape, subsref updated.
  • Support functions test_equal, zeroslike and isactive added.

  • Printout removed when optPar(20) was set for filterSQP.

  • Help updated.

Sep 14th 2005

TOMLAB Base Module v4.9 released:
  • glbDirect problem with NaN values fixed.
  • Demo improvements for TOMLAB /LDO.
  • One more MIQP problem added to distribution.
  • One more LLS problem added to tomlab\testprob. LSQLIN fails on this case.
  • cls_prob, con_prob updated.
  • slsSolve much faster with SNOPT7 as a subsolver - CG used instead for optPar(66).
  • Safeguarded starting point for PDCO modified.
  • COLMMD replaced by COLAMD in ComputeQR, ssqls, cne, csne, colamd and qls.
  • Example updated for expSolve.
  • chs_prob updated (Hessian fixed for test case 104).
  • isstr replaces by ischar in applicable files (30).
  • Two new algorithms available in clsSolve - Li-Fukushima MBFGS and Broydens method.
  • expSolve now takes eType as a new input.
  • Improved error messages if problems with user routines.
  • Weighting for least squares problems updated - numerical differentiation caused double weights.
  • Automatic differentiation for least squares problems updated.
  • Help updated for TLSQR.
  • 'help solvername' will now display suitable information for most solvers.
  • New quick guide released - detailed information about patterns, derivatives and warm start provided.
  • BINTPROG added to optimization toolbox interface.
  • L1LinSolve and L1Solve fully sparsified. Code more efficient now.
  • infSolve - patterns now always sparse.
  • goalSolve - improved error messages and sparsified code.
  • slsSolve - code improved for speed.
  • miqpAssign, miqqAssign and minlpAssign updated.
  • Number of Trials in estConsPattern and estJacPattern reduced to 2.
  • estd2cPattern added to distribution (needed when estimating d2LPattern).

  • optPar 55-62 removed from the solvers. Restarts need to be done directly from TOMLAB.
  • LSSOL printing problem removed for problems with no linear constraints.
  • LPOPT/LP-MINOS/QP-MINOS/QPOPT/SNOPT/SNOPT7/SQOPT/SQOPT7 now accepting problems without linear constraints.
  • StateDef revised for SNOPT 6.
  • All version of SNOPT using nonderivative line search when Prob.CheckNaN is set.
  • LSSOL, MINOS, LP-MINOS, QP-MINOS, SNOPT all have updated evaluation counters.

  • Problem with demo license fixed.

  • Try-catch statement added to identify installation problems.

  • Major update. Issue resolved regarding defaults (correct defaults provided at all times).
  • filterSQP constraint evaluations corrected.
  • Memory problem resolved for filterSQP (solving a small problem after a large could cause problems).

  • Major update. Defaults now specified correctly.

  • Return code revision. The solver sometimes returned feasibility for infeasible problems.

  • isscalar added to functions supported.

  • Default time limit changed to 1e7.

  • Updated cpxtomtest1 to correct format.

  • Updated handling of d2LPattern. Solver should perform better when all parts are analytical.
  • d2LPattern automatically set for LP and QP problems.
  • HessPattern automatically set for problems with linear or quadratic objectives.

  • Hessian and d2c handling improved.
  • Constraints only calculated if needed.
  • Working correctly for all QP problems.
  • No size restrictions to demo version anymore.
  • d2LPattern automatically set for LP and QP problems.
  • HessPattern automatically set for problems with linear or quadratic objectives.

Jul 8th 2005

TOMLAB Base Module v4.8 released:
  • SIGNIFICANT SPEED IMPROVEMENTS (TOMLAB /SOL) - solvers doing (non-costly) callbacks to user routines now 20-50 % faster.
  • preSolve updated to handle single constraints.
  • M-file help improved system-wide.
  • Manual more detailed about user supplied parameters.
  • LPSOLVE renamed to MILPSOLVE.
  • Optimization toolbox interface updated to handle more cases.
  • Contents.m files now present in all folders.
  • PrintResult working better when there are several solutions.
  • DualSolve updated to handle changed version of lpSolve (now lpSimplex).
  • Safe-guarded starting point added to mipSolve.
  • Geometric programming added as a general problem type.
  • As a result of Matlab Compiler issues, pragma %#mex added to all solvers with m-file help for dll/mex.
  • lpconAssign and qpconAssign modified to standard notation format.
  • llsAssign safe-guarded for y and t input - they are now automatically converted to full vectors.
  • Geometric programming problem added to the TOMLAB Quickguide.
  • MAD examples added to TOMLAB Quickguide.

  • rbfSolve and ego has improved initial strategies.

  • Specs file functionality updated.
  • Minor bug related to LSSOL and NLSSOL fixed.

  • No longer possible to use general TOMLAB parameters to set options.
  • The optPar vector needs to be used for all alterations.
  • Full manual review of control parameters.

  • Return code and ExitFlags updated for SNOPT7 and SQOPT7.
  • New parameters made available for SNOPT and SQOPT.
  • Several parameter and manual updates for all SNOPT version (6 and 7).
  • SNOPT 6 has control vector (optPar) of length 65, SNOPT 7 of length 71.
  • SNOPT 6 information:
        Correction of defaults for optPar 2,10,27,28,29,34,35,36,42
        New parameter LU SWAP TOLERANCE as optPar(25), LU DENSITY removed.
        New parameter PROXIMAL POINT METHOD as optPar(64)
        New parameter PENALTY PARAMETER as optPar(65)
        New parameter NEW SUPERBASICS (MINOR SUPERBASICS) as optPar(66)
        optPar(44): FEASIBLE EXIT (Obsolete)
        Help added comments about parameters only available using SPECS file.
  • SNOPT 7 information:
        Correction of defaults for optPar 2,10,27,28,29,34,35,36,42
        New 6.2 parameter LU SWAP TOLERANCE as optPar(25), skip LU DENSITY
        New 6.2 parameter PROXIMAL POINT METHOD as optPar(64)
        New 6.2 parameter PENALTY PARAMETER as optPar(65)
        New 6.2 parameter NEW SUPERBASICS (MINOR SUPERBASICS) as optPar(66)
        optPar(44): FEASIBLE EXIT (Obsolete)
        Adding comments about parameters only available using SPECS
        New 7.1 parameter: optPar(66). QPSOLVER CHOLESKY,CG or QNCG
        New 7.1 parameters: optPar(67): CG TOLERANCE, optPar(68): CG ITERATIONS
        New 7.1 parameter: HESSIAN (or CG) PRECONDITIONING, optPar(69).
        New 7.1 parameter: SUBSPACE, optPar(70).
        New 7.1 parameter: optPar(71): HESSIAN DIMENSION (also REDUCED HESSIAN)
        Removed special setting of SUPERBASICS (optPar(48)).

  • Control vector (optPar) now of length 71.
  • Added new parameters AIJ CONVERGENCE, SUBSPACE, LU WRAP.
  • Change defaults for #6,42,43 (optPar control vector).
  • Increased SUPERBASICS, optPar(48) to avoid too low value.
  • LP-MINOS and QP-MINOS updated with new parameters.

  • Version 9.1 released. See the manual for more information.
  • Possible to warm start solution process for MILP and MIQP problems.
  • New local branching heuristics.
  • Issues warning of quadratic problem (hessian) is not symmetric.
  • Possible to supply starting vector for mixed integer problems.
  • Problems corrected for sparse matrix handling in MATLAB R12 and R13 (only applies to problems with quadratic constraints).

  • Warning messages removed.

  • Safe-guarded in Jacobian calculation.
  • Both minimization and maximization problems possible to solve.

  • Updated to choose appropriate HESSOPT depending on user routines supplied.

  • Released.
  • GP is an interior-point package for geometric programming. Problems which are non-differentiable in the optimum can be efficiently solved.

  • Supported for Linux 64-bit and MAC OS X.

  • Now supported for Sun Solaris.

  • Now supported for all operating systems.

May 11th 2005
Tomlab Optimization announces an alliance with National Instruments. TOMLAB to be released for LabVIEW users during the spring of 2006.

The primary modules to be developed are:
  • TOMLAB Base Module - general interface code and tools. Solvers for linear, mixed-integer programming and global optimization.
  • TOMLAB /FITTING - a package for linear and nonlinear fitting problems.
  • TOMLAB /LOCAL - state-of-the-art solver package for local optimization.
  • TOMLAB /GLOBAL - package for large-scale global optimization.
  • TOMLAB /CGO - costly (expensive) global optimization tools.
Development and deployment packages will be offered for all options.

Please contact for more information.

May 9th 2005

TOMLAB Base Module v4.7 released:
  • Example for calling TOMLAB solutions from Excel included.
  • More standalone examples included in the distribution. An Excel example is now included.
  • Minor name changes to avoid conflicts with other toolboxes:
        printmat -> PrintMatrix
        assign -> ShowAssignment
        simplex -> ShowSimplex
        aic -> AkaikeIC
        runfleq1 -> Trunfleq1
  • New folder tomlab/common includes all files common to any installation.
  • Only one startup file included in the distribution, all others have solver specific names.
  • Linear/Quadratic problem with nonlinear constraints added to quick guide.
  • Simulation problem added to quick guide.
  • Manual updated to clarify user parameters in Prob.
  • uhs_prob, chs_prob and uc_prob updated.
  • checkFuncs added. We recommend that this routine is used to check TOMLAB problem before execution.
        Prob = *Assign();
        Prob.user.a = a;
  • chs_prob, con_prob, glc_prob, minlp_prob linear constraints updated.
  • Makeinitfile modified to avoid print-outs to screen.
  • glcCluster updated to accept maxFunc3 input directly.
  • glcFast updated to handle all variables fixed.
  • New features for sending user parameters to Init Files, see Section 14.2 in the general TOMLAB manual.
  • Compatibility verified with MATLAB R14 SP2.
  • Review for MATLAB R12. CPLEX, PENSDP, PENBMI, SOCS, BARNLP, SPRNLP now stable.
  • Problem with Prob.NumDiff = 4 resolved. The spline is now converted correctly.
  • A dedicated version for Linux on MATLAB R14 released.
  • glcFast safe-guarded for user routines with different number of inputs.
  • lpSolve renamed to lpSimplex.
  • LPSOLVE (LP and MILP solver) added to the TOMLAB Base Module. See and TOMLAB manual for more information.
  • 19 more glb_prob's added to the distribution.
  • New version of glcSolve released.
  • glbDirect and glcDirect released (will replace glbFast and glcFast eventually).
  • Printing error in nlp_r corrected.
  • L1LinSolve revised to avoid an intermediate LS structure (improved memory handling).
  • TOMLAB/MATLAB engine call from C program included in the distribution.
  • Further development of a constrained mixed-integer DIRECT algorithm has been implemented in the new releases of glcSolve and glcDirect.

  • Updated to better handle return codes from glcFast.
  • Improved handling of fixed variables.
  • Possible to run 1-dimensional integer problems efficiently.
  • Checks on duplicate points added.
  • Stops if all integer combinations tried.

  • A full manual is now available from the manuals page.

  • Version 7 of SNOPT and SQOPT is now available.
  • SNOPT 7 is called with tomRun('snopt7'...)
  • Significant performance improvements - much larger problems can now be solved.
  • Documentation updated for scaling, 2 default for LP, 0 for NLP (good to try scaling for NLP's manually).
  • SQOPT documentation updated for scaling, 2 default for LP, 0 for QP (good to try scaling for QP's manually).

  • Examples aircrew, biptest, iptest, tomtest1, tomtest2 have new names, cpx prepended.
  • Documentation updated for sensitivity analysis - possible to set indices directly rather than the start and end points.
  • New options added to the solver, guided dives and more.

TOMLAB /Xpress:
  • Examples aircrew, biptest, iptest, tomtest1, tomtest2 have new names, xp prepended.

  • New manual available.

  • Released. A new manual is available from our manuals page.

  • Now possible to turn off screen printing.

  • Print out bug removed for feasibility tolerances.
  • Works for pure integer/discrete programming problems.
  • Feasibility problems solved.
  • New manual available.
  • Defaults now different for DISTANCE_FACTOR.
  • More stable release - several issues fixed.

  • Bug in multiplication removed.

  • NLSSOL safe-guarded for incorrect length in residuals.
  • Constraint handling updated for NLSSOL.

Mar 11th 2005

64-bit Linux platform now supported:
  • Extremely large problem can now be solved - with enough memory a complete new range of applications have been enabled.
  • The supported solvers packages are: TOMLAB Base Module, TOMLAB /SOL (/SNOPT, /NPSOL, /MINOS), TOMLAB /CPLEX, TOMLAB /MSNLP, TOMLAB /LGO, TOMLAB /CGO, TOMLAB /MAD.

Feb 2nd 2005

TOMLAB Base Module v4.6 released:
  • Bug in nlpSolve removed. Convergence testing was incorrect.
  • Penalty values in conSolve now dampened to promote convergence.
  • conSolve now works better for unconstrained problems.
  • Major updates for goalSolve, several new subsolvers are now working. Better handling when no analytical derivatives are available.
  • glcCluster has slightly modified input fields and better defaults.
  • Pattern checks added to checkDerivs.
  • chs_prob set extensively corrected.
  • Major revisions of L1Solve, infSolve and slsSolve.
  • PDCO and PDSCO updated.
  • WeightType 3 removed from LSEI.
  • Tfzero and Tfmin conformed to TOMLAB standard.
  • Complete print file review of the entire system. Minimal file printing by default from all solvers.
  • Variable, linear/nonlinear constraint states revised for all solvers.
  • lpDemo and qpDemo updated to new TOMLAB format.
  • Several updates to LDO to conform to TOMLAB structure.
  • All assign routines does length checking on x_0, x_L and x_U.
  • Starting point safe-guarded in all solver packages.
  • tomSolve updated with more solver options.
  • fmincon interface (drop-in replacement) updated to handle more cases.
  • Numerical differentiation for NLLS problems improved.

  • Dual CPU activated in demo license. Please contact for more information.
  • Possible to warm start the solution process for LP problems.

  • Version 2.0 released.
  • Safeguarded for missing initial vector.

  • Version 2.0 released.

  • New manuals for most solvers (MINOS, SNOPT, QPOPT, SQOPT).

  • Default for number of superbasics is automatically set to better number.
  • Safeguard added in SQOPT for dense QP problems.

  • WeightType 3 removed from LSSOL.
  • Print files secured in the code to avoid unexpected behavior.

  • Minor interface upgrades.
  • Stability improvements.

  • Print files no longer generated by default.

  • Version 3.0 released.
  • Direct access to MSNLP from the interface.
  • The embedded local solver LSGRG-2 can be executed independently.
  • Empty print file removed.
  • More detailed algorithmic description available in the TOMLAB /OQNLP manual.
  • Possible to generate a LOCALS file with all local solutions.
  • Return codes reviewed.

  • WarmDefDUNDEE updated with estimated problem patterns.
  • New definition of infinity for better performance.
  • New TOMLAB /MINLP manual with algorithmic description available for download.

  • Variable and linear constraint states added to output.

  • New version with improved stability.
  • Solver options are now set in Prob.LGO.options.

  • Solver options are now set in Prob.CONOPT.options.

  • Formal release.
  • A manual and quickguide are available from the manuals page.

Dec 17th 2004

TOMLAB Base Module v4.5 released:
  • WarmDefDUNDEE added for the TOMLAB /MINLP solvers.
  • Direct access to the tomsol.dll enabled - improving speed for matrix - vector multiplications in MATLAB/TOMLAB. tomsol(0) will display the help.
  • PrintAssign working with user given names for problems.
  • 30 additional test cases for MILP problems included in the distribution. See mip_prob.m
  • 2 additional MIQP problems added.
  • 42 additional LP problems added.
  • 27 extra QP problems added.
  • lpconAssign and qpconAssign added for LP and QP problems with nonlinear constraints.
  • xnargin modified provide better error messages.
  • xxx_prob.m removed from distribution and integrated in applicable files.
  • Bugs in estHessPattern/estConsPattern/estJacPattern fixed.
  • fmincon interface updated.
  • tomRun changed to tomRun('solver', Prob, PriLev), i.e. no need to give a third [] parameter.
  • TLSQR updated to exit more smoothly on linear least squares problems with equality constraints.
  • mipSolve and cutplane can now solve LP problems.
  • WARNING - glcAssign has been changed, setupFile, nProblem and KNAPSACK has been removed as inputs - WARNING

MATLAB Compiler:
  • C/C++ examples for Windows/Linux/Sun included in the general distribution. This exemplifies how to call TOMLAB in standalone mode.
  • An updated TOMLAB /SAL guide is available from the download page.

TOMLAB /KNITRO v4.0 released:
  • Many new features, including a new solver algorithm option.
  • Defaults changes to better suit the user. If analytical gradient, then finite-difference Hessian-vector products are used. If only the objective function is given, a memory limited quasi-Newton Hessian is used.
  • A few minor fixes included in the new version.
  • Download the latest manual for more information.

  • New defaults for line search when using numerical derivatives.
  • Updates to handling of defaults for the solver settings.

TOMLAB /CGO v2.5 released:
  • rbfSolve updated for better handling of binary and integer variables.
  • EGO improved to work better with bad input data.
  • New manual with detailed algorithmic descriptions for rbfSolve.
  • A problem with rbfSolve and ego sampling outside the bounds for costly integer programming has been removed.
  • Safeguards added when local solver cannot return integer solution.

TOMLAB /XA v14 released:
  • New features for IIS.
  • Barrier works for LP problems.
  • Faster simplex and MIP solutions.

  • MaxCPU added.
  • Minor printing bugs removed.

  • (X)MPS and other input files can now be read into MATLAB. LP, MILP, QP and MIQP supported.
  • The AMPL format is supported in TOMLAB /AMPL.

  • New version of mtimes_dab.
  • ceil and cumprod added to the distribution.

  • Printlevel added to amplAssign.
  • Safeguard for cases with no nonlinear constraints.

  • New interface routine included which avoids numerical differentiation for integer variables.

  • Minor printing fixes.

  • Released for Linux.

TOMLAB no longer supported for MAC OS 9. Existing licenses can be transferred to MAC OS X at no cost.

Sep 28th 2004
Infeasibility analysis added to TOMLAB /CPLEX and TOMLAB /Xpress:
Both packages will now generate an optional IIS (irreducibly inconsistent set) when the problem is infeasible. This information can be used to remove or identify restricting parts of the models analyzed.

Sensitivity analysis now in TOMLAB /CPLEX and TOMLAB /Xpress:
The new functionality will detail how much a model can be modified without affecting the solution obtained. It is possible to study changes in the objective function, right-hand-side vector and bounds.

MATLAB Compiler:
A complete revision of TOMLAB for users of the MATLAB Compiler. Compiler pragmas have been added to most TOMLAB functions to simplify the use. mFiles is now called tomFiles due to a potential conflict.

General TOMLAB upgrades:
  • miqqAssign added for mixed-integer problems with quadratic constraints.
  • User routine - checkDerivs added. The user may now verify all supplied routines using the TOMLAB format.
  • TOMLAB /CGO /XA /CPLEX /Xpress - minor bug fixes and improvements for better usability.
  • TOMLAB /MAD v1.2 released.
  • Compatibility confirmed for MATLAB 7.

Jul 2nd 2004
TOMLAB SAL User's Guide available for customers building embedded solutions. The guide details the preferred procedures and provides examples.

May 21th 2004
TOMLAB Quickguide available for all problem types. Download from the manuals page. Examples for LP, MILP, QP, MIQP, MIQQ, NLP, MINLP, LLS, NLLS, GLB, GLC, SDP, BMI, MINIMAX, L1, GOAL.

May 18th 2004
TOMLAB /OQNLP v2.0 released. Many general improvements to the multi-start features. Changes to the distance and merit filters make them theoretically sounder and more adaptive to the problem instance. Stochastic drivers are implemented whose performance is comparable to the OptQuest scatter search implementation, if reasonably tight bounds on all variables are imposed.

Apr 28th 2004
TOMLAB /PATH released. Linear and nonlinear mixed complementarity problems can now be solved with TOMLAB. The new assign routines lcpAssign and mcpAssign should be used to create the problem. TOMLAB /PATH also handles linear and convex quadratic programming problems. Developed in cooperation with University of Wisconsin at Madison.

Apr 17th 2004
The TOMLAB Base Module now features several new additions for derivative handling.
  • Automatic estimation of ConsPattern, HessPattern and JacPattern for large-scale problems.
  • New method for numerical differentiation using standard MATLAB splines.
  • The TOMLAB interface routines now supply the user with row, column and variable information for use in their code. With this information, only segments of costly code need to be computed.
  • Many new features does automatic problem validation. Element sizes in linear constraints can be checked, the PreSolve capabilities have been improved, a special flag for warning messages is now used.
  • A MaxCPU flag has been introduced for several solvers, limiting the amount of CPU time consumed before returning to the command line.
TOMLAB /CPLEX has been updated with a special network interface. The user can now specify a set of nodes and arcs when solving their problems.

Mar 25th 2004
TOMLAB released for MAC OS X users. TOMLAB Base Module, TOMLAB /SOL (MINOS, NPSOL, SNOPT) and TOMLAB /MAD. More packages are available but evaluations should be performed on another platform.

Mar 22nd 2004
TOMLAB /CGO v2.0 released. rbfSolve and EGO now handle integer variables. TOMLAB /OQNLP and glcCluster can be used as a sub solver.

Mar 8th 2004
Tomlab announces a cooperation with LINDO Systems. A complete integration of the LINDO API will be made available in TOMLAB. Read more >>

Mar 7th 2004
TOMLAB /MAD released. A package for MATLAB Automatic Differentiation. The package is fully integrated in the TOMLAB optimization environment. Read more >>

Feb 26th 2004
TOMLAB /NLPQL released. A package for multi criteria optimization and nonlinear fitting problems. Read more >>

Feb 4th 2004
TOMLAB v4.2 released. The new distribution of TOMLAB has gone through a major upgrade.

List of upgrades:
  • TOMLAB Base Module now features fully integrated support for MAD (MATLAB Automatic Differentiation). The user simply has to specify two flags in the standard Prob structure to use MAD. TOMLAB /MAD released shortly.
  • All TOMLAB solver are now directly callable, Prob = *Assign( ... );, Prob = ProbCheck( ... );, Result = solverTL(Prob);
  • A complete system revision has further integrated TOMLAB with the driver routine tomRun.
  • New version of TOMLAB /CPLEX v9.0. Explicit quadratic constraints are now supported. With a special license the solver may be executed on up to 64 parallel processors.
  • TOMLAB /CGO v1.6 released. Better handling of non-costly nonlinear constraints implemented.
  • TOMLAB /MINOS now support LU Rook Pivoting.
  • TOMLAB /SNOPT now support LU Rook Pivoting, and LU Diagonal Pivoting.
  • Upgraded versions of glbSolve, glbFast, glcSolve, glcFast, glcCluster, PDCO and Tlsqr.
  • Minor bug fixes for TOMLAB /XA.

Jan 28th 2004
TOMLAB /LGO for global and convex nonlinear optimization is now available. Read more >>

Jan 28th 2004
A collection of global optimization test models is available in the TOMLAB distribution. The test collection is located in tomlab\testprob, in files lgo1_prob and lgo2_prob, after installation. Reference: Pintér, J.D., Bagirov, A., and Zhang, J. (2003) An Illustrated Collection of Global Optimization Test Problems. Research Report, Pintér Consulting Services, Inc. Halifax, NS, Canada; and CIAO-ITMS, University of Ballarat, Ballarat, Vic., Australia. Download TOMLAB v4.2

Nov 27th 2003
TOMLAB /XA is now available for download.

Nov 21st 2003
TOMLAB v4.1.1 released. Some bug fixes. New versions of TOMLAB /CPLEX, TOMLAB /KNITRO, TOMLAB /PENSDP and TOMLAB /PENBMI available. Download TOMLAB v4.1.1.

Oct 10th 2003
Tomlab announces a cooperation with Sunset Software Technology. Sunset's XA solver package is now available to the MATLAB user. TOMLAB /XA features several advanced options for large-scale linear, mixed-integer and quadratic programming. Please refer to our TOMLAB /XA product summary.

Sept 24th 2003
Tomlab Optimization announces a cooperation with Boeing Phantom Works. Tomlab Optimization will develop MEX interfaces to Boeing's SOCS (The Sparse Optimal Control Software Family) for use in MATLAB.

SOCS is a general-purpose software package for solving optimal control problems. Applications include trajectory optimization, chemical process control and machine tool path definition. SOCS also includes solvers for Large Scale NLP + Differential-Algebraic Equations. This capability includes mesh refinement, sparsity detection and sparse finite difference software.

Sparse Nonlinear Programming exploits state-of-the-art sparse linear algebra technology to solve very large optimization problems orders of magnitude faster than traditional methods. Applications with more than 100,000 variables and constraints can now be solved efficiently on desktop computers. Sparse SQP, and sparse primal-dual interior point, with special interfaces for least squares, LP, and QP applications.

Sept 9th 2003
Tomlab Optimization presents a paper at the Nordic MATLAB Conference 2003 (21-22 Oct). Download the abstract here.

Sept 5th 2003
TOMLAB v4.1 released. Some bug fixes; Dual gap is now handled correctly in mipSolve. With this new version three more products are released, TOMLAB /KNITRO v3.0, TOMLAB /OQNLP v1.0, TOMLAB /CONOPT v3.1. TOMLAB v4.1 can be downloaded here.

Sept 4th 2003
New User's Guides for TOMLAB /MINOS, TOMLAB /NPSOL, TOMLAB /SNOPT, TOMLAB /SOL and TOMLAB /PENOPT has been released. See all manuals >>

Sept 2nd 2003
The TOMLAB homepage has been slightly changed and has some new features.

Aug 30th 2003
The TOMLAB User's Guide has been updated. New User's Guides for TOMLAB /Xpress, TOMLAB /CPLEX and TOMLAB /AMPL has been released. See all manuals >>

Aug 12th 2003
TOMLAB /AMPL v2.0 released. Now problems formulated in AMPL can be run in TOMLAB. You can read more about the product at

May 26th 2003
TOMLAB v4.0.6 released. New features and fixes include:
  • Special handling of simulation problems, where it is necessary to compute both function and constraints in one call. Tomlab now also efficiently estimates unknown gradients for both function and constraints.
    New routines simAssign, and related gateway functions: sim_fc, sim_gdc, sim_f, sim_g, sim_c, sim_dc handle the new feature.
  • Fixes in problem assignment routines for increased stability
  • Revised constraint handling in rbfSolve and ego
  • General bug fixes

May 15th 2003
Tomlab Optimization's website has a new look!

February 28th 2003
TOMLAB v4.0.5 released. Minor bug fixes.

February 3rd 2003
TOMLAB v4.0.1 released. TOMLAB's two input formats now take function_handle input, minor bug fixes.

January 31st 2003
TOMLAB v4.0 released. Now fully compatible interface to Optimization Toolbox, with selectable Tomlab solver. Tomlab automatically estimates second order information for TOMLAB /MINLP if needed, using numerical differences and given sparsity patterns. Partial information in derivatives possible with new CheckNaN option. TOMLAB /SOL v4.0 with new versions of SNOPT and SQOPT. New version of TOMLAB /PENSDP. Improvements in mipSolve, preSolve and general speed improvements.

January 31st 2003
TOMLAB /PENBMI v1.0 released, solves semidefinite programming with bilinear matrix constraints (BMI). The toolbox is developed in cooperation with PENOPT Gbr and researchers Michal Kocvara and Michael Stingl, University of Erlangen-Nürnberg.

Jan 4th 2003
TOMLAB /MINLP v1.0 is released, solving mixed-integer nonlinear programming problems, NLP, QP and mixed-integer QP problems. The toolbox is developed in cooperation with Roger Fletcher and Sven Leyffer, University of Dundee.

Oct 21st 2002
TOMLAB /CGO 1.4 now available for download!

Oct 11th 2002
TOMLAB /CPLEX, based on CPLEX 8.0, is released, solving mixed-integer linear and quadratic programming problems, and large-scale linear and quadratic problems. The toolbox is developed in cooperation with ILOG Inc.

Oct 11th 2002
TOMLAB /SOL v3.2.2 and /CGO 1.3 now available for download!

Aug 30th 2002
TOMLAB /PENSDP v1.0 released, solves semidefinite programming and LMI problems. The toolbox is developed in cooperation with PENOPT Gbr and researchers Michal Kocvara and Michael Stingl, University of Erlangen-Nürnberg.

Aug 30th 2002
TOMLAB /SOL v3.2.1, /CGO 1.2 and /PENSDP v1.0 now available for download!

July 5th 2002
TOMLAB /SOL v3.2. Speed improvements. New types of optimization problems treated by Tomlab: mixed-integer quadratic and nonlinear programming, semidefinite programming. New license handling enabling increased number of Tomlab solver toolbox options.

April 14th 2002
TOMLAB /SOL v3.1.2
  • New solvers for Minimax, L1 data fitting and sparse nonlinear least squares. All explicitly treating linear and nonlinear constraints.
  • New versions of SNOPT and MINOS.
  • Improvements in global optimization solvers, now handling explicitly linear and nonlinear constraints.
  • DIRECT algorithm in glcSolve/glcFast expanded to handle linear and nonlinear equality constraints.

July 1st 2002
TOMLAB /Xpress v2.3.2 now runs on Windows, Linux, HP and Sun systems.
  • Features MIQP solver, e.g. suitable for portfolio optimization.
  • Available for download with Limited Edition license
  • A demo license without size restrictions is available on request.

Jan 11th 2002
Compile Tomlab to standalone with MCC.
With special Tomlab standalone license, create your embedded system with state-of-the-art optimization.

Jan 1st 2002
Tomlab Optimization opens office in USA.

Jan 11th 2002
Latest Development in Global Optimization
  • New global optimization solver glcCluster , a combination of glcFast, a clustering algorithm, and a local solver.
    Solves nonconvex problems with higher dimensions faster and more robust.
  • New fast constrained mixed-integer global optimization solver glcFast , a MEX file Fortran version of the constrained DIRECT algorithm in glcSolve , See the description of glbSolve for some test results with glcFast.
  • New fast global optimization solver glbFast , a MEX file Fortran version of the DIRECT algorithm in glbSolve. See the description of glbSolve for some test results with glbFast.

Jan 11th 2002
New toolbox for Costly Global Optimization - TOMLAB /CGO

    Tomlab © 1989-2022. All rights reserved.    Last updated: Oct 3, 2022. Site map. Privacy Policy