Skip to content

Latest commit

 

History

History
294 lines (210 loc) · 9.08 KB

CHANGES.md

File metadata and controls

294 lines (210 loc) · 9.08 KB

Change history for MIPS

Version 1.5.1 - May 10, 2024

5/10/24

  • Release 1.5.1.

4/16/24

  • Add full_hist option to enable saving in output.hist trajectories of x, z, g, h, lam, and mu (issue #4).

3/12/24

Version 1.5 - Dec 12, 2022

12/12/22

  • Release 1.5.

10/28/22

  • Add to mplinsolve() the ability to return a struct containing the matrix LU factorization, and to reuse this pre-factored matrix to solve additional systems with different right-hand-sides by passing the struct in place of the A matrix to subsequent calls.
  • Add option to mplinsolve() to solve transposed systems by setting opt.tr to 1, including when providing the pre-factored matrix for the original, non-transposed system.

Version 1.4 - Oct 8, 2020

10/8/20

  • Release 1.4.

9/23/20

  • Add feature detection functions for LU and PARDISO, defining tags 'lu_vec', 'pardiso_legacy', 'pardiso_object' and 'pardiso' for have_feature(). Requires MP-Test 7.1 or later.
  • Remove custom LU and PARDISO detection code from mplinsolve and test functions, and switch to using have_feature().

8/20/20

  • In User's Manual, add missing linsolver option in Table 3-3.
  • In User's Manual, add Sections 4 Other Functions, 4.1 Linear System Solvers - mplinsolve, and move Section 3.3 Quadratic Programming Solver to 4.2.

8/18/20

  • Silence inadvertent output from mplinsolve() when called without solver input argument.
  • Fix fatal errors when mplinsolve() is called with LU solver and dense A matrix.

4/23/20

  • Remove additional have_fcn() dependencies in mips(), t_mips_pardiso() and t_qps_mips().
  • INCOMPATIBLE CHANGE: Calling mips() with opt.linsolver set to 'PARDISO' now results in a fatal error if PARDISO is not installed, rather than warning and continuing with the default linear solver.

8/15/19

  • Fix bug in initialization of mu that resulted in fatal error in t_mips for some versions of MATLAB (e.g. R2013b on Mac).

Version 1.3.1 - Jun 20, 2019

6/20/19

  • Release 1.3.1.
  • Add CITATION file.
  • Other miscellaneous documentation updates, e.g. MATPOWER website links updated to https://matpower.org.

Version 1.3 - Oct 30, 2018

10/30/18

  • Release 1.3.

9/10/18

  • Add support for PARDISO v6.x.

5/7/18

  • Ignore warnings from Octave 4.4 about calling lu with less than 4 outputs and a sparse input.

11/29/17

  • Fix typo preventing pardiso.dparm options from being set.
  • Fix incorrect use of PARDISO options in t_mplinsolve().

3/16/17

  • Move mplinsolve PARDISO options to opt.pardiso in preparation for adding options for other solvers.
  • Add mplinsolve solver option 'LU' for explicit LU decomposition with back substitution, with options in opt.lu for specifying the number of output arguments in call to lu (opt.lu.nout), whether to use permutation vectors or matrices (opt.lu.vec) and pivot threshold options (opt.lu.thresh). The following values for the solver argument act as shortcuts for specifying various combinations of options: 'LU3', 'LU3a', 'LU4', 'LU5', 'LU3m', 'LU3am', 'LU4m', 'LU5m'. See help mplinsolve for details. Thanks to Jose Luis Marin.

12/21/16

  • Add Travis-CI integration. Thanks to Richard Lincoln.

Version 1.2.2 - Dec 16, 2016 (released with MATPOWER 6.0)

12/15/16

12/9/16

  • Renamed from MATLAB Interior Point Solver to MATPOWER Interior Point Solver.

12/6/16

  • Remove dependence of t_mpsolve() on presence of have_fcn() to detect PARDISO installation.

11/1/16

  • Released MATPOWER 6.0b2.

Version 1.2.1 - Jun 1, 2016 (released with MATPOWER 6.0b1)

3/27/15

  • Fixed issue where default value of feastol option was not being set correctly in mips() when called directly (or via qps_mips()) with feastol = 0.

Version 1.2 - Mar 20, 2015 (released with MATPOWER 5.1)

3/19/15

  • Added support for using PARDISO as linear solver for computing interior-point update steps in MIPS, via new mplinsolver() function and linsolver option.

2/25/15

  • Switch to more permissive 3-clause BSD license from GPL 3.0.

Version 1.1 - Dec 17, 2014 (released with MATPOWER 5.0)

12/2/14

  • Additional user-settable options: xi, sigma, z0, alpha_min, rho_min, rho_max, mu_threshold and max_stepsize.
  • INCOMPATIBLE CHANGE: The name of the option to mips() to specify the maximum number of step-size reductions when step_control is on was changed from max_red to sc.red_it for consistency with other MATPOWER options.

Version 1.0.2 - Nov 5, 2013 (released with MATPOWER 5.0b1, Jul 1, 2014)

11/5/13

  • Fixed a bug in MIPS where a near-singular matrix could produce an extremely large Newton step, resulting in incorrectly satisfying the relative feasibility criterion for successful termination.

Version 1.0.1 - Apr 30, 2012

4/30/12

  • Fixed fatal bug in MIPS for unconstrained, scalar problems. Thanks to Han Na Gwon.

12/14/11

  • Released MATPOWER 4.1.

Version 1.0 - Feb 7, 2011 (released with MATPOWER 4.0)

  • no change

Version 1.0b5 - Dec 13, 2010 (released with MATPOWER 4.0b5)

  • no change

Version 1.0b4 - May 21, 2010 (released with MATPOWER 4.0b4)

5/10/10

  • Modified input args for Hessian evaluation function for MIPS. Requires cost_mult as 3rd argument.

4/27/10

  • Check for NaN's in update step.

Version 1.0b3 - Apr 19, 2010 (released with MATPOWER 4.0b3)

4/19/10

  • Changed licensing to GNU General Public license. See LICENSE and COPYING files for details.

4/6/10

  • GNU Octave compatibility! (tested with Octave 3.2.3)

Version 1.0b2 - Mar 19, 2010 (released with MATPOWER 4.0b2)

3/10/10

  • Added optional input arg to mipsver() function to trigger return of entire version struct with Name, Version, Release and Date (similar to MATLAB's ver() function).
  • Massive help text update to more closely match MathWorks conventions; function names in ALL CAPS, See also ..., Examples, etc.

3/8/10

  • Added mipsver().

3/1/10

  • Added a version number and printing of MIPS version lines to verbose output.

2/25/10

  • Swapped g and h (G and H) in notation to match convention used in previously published stuff.

1/25/10

  • Added qps_mips(), with calling syntax similar to quadprog() from the Optimization Toolbox. The main difference from the quadprog() API is that the constraints are specified asl <= A*x <= u, instead of A*x <= b and Aeq*x == beq.

1/19/10

  • Renamed the pure-MATLAB interior point solver from PDIPM to MIPS (MATLAB Interior Point Solver).

1/18/10

  • Changed order of input args to pdipm(), added option for single input struct (like fmincon), more documentation, all constraints are now optional, returns exitflag = -1 for numerically failed, output includes message field, lambda only includes relevant fields. Added tests for pdipm as standalone solver.

1/12/10

  • Added history field to the output with trajectories of obj, termination criterion, etc.

1/10/10

  • Added acknowledgement of port from Hongye Wang's code.

Version 1.0b1 - Dec 24, 2009 (released with MATPOWER 4.0b1)

11/4/09

  • Removed unnecessary return statement at end of all M-files. If anything it should be an end statement, but even that is optional, so we just let functions get terminated by the end-of-file or another function declaration.

6/5/09

  • Break out of algorithm is any element of x becomes NaN.

3/24/09

  • Added step-controlled PDIPM variant.

3/19/09

  • Fixed some bugs in default args.

3/13/09

  • Added a pure MATLAB implementation of the PDIPM (primal-dual interior point method) solver.