MP-Opt-Model 4.2

What's New in MP-Opt-Model 4.2

Released May 10, 2024

Below is a summary of the changes since version 4.1 of MP-Opt-Model. See
the CHANGES.md file for all the gory details. For release notes
for previous versions, see Appendix C of the MP-Opt-Model User's
Manual.

New Features:

• Option for opt_model.add_lin_constraint() to provide/store the
  transpose of the A matrix instead of the original. This can
  potentially save significant memory for sparse matrices with many
  more columns than rows. E.g. storage constraints in MOST
  for 8760 hour planning horizon.
• Add support to nlps_master() for calling nlps_<my_solver>() by
  setting opt.alg to '<MY_SOLVER>' to allow for handling custom
  NLP solvers.
• Add support to miqps_master() for calling miqps_<my_solver>()
  by setting opt.alg to '<MY_SOLVER>' to allow for handling custom
  MILP/MIQP solvers.
• Add to the parse_soln() method of opt_model an optional stash
  input argument that, if present and true, causes the parsed
  solution to be stored back in the object, as the solve() method
  was already doing when opt.parse_soln is true.
• New Sphinx-based Reference documentation.
• New functions:
  • convert_lin_constraint() converts linear constraints from a
    single set of doubly-bounded inequality constraints to separate
    sets of equality and upper-bounded inequality constraints.
  • convert_lin_constraint_multipliers() converts multipliers on
    linear constraints from separate sets for equality and
    upper-bounded inequality constraints to those for doubly-bounded
    inequality constraints.
• New opt_model methods:
  • is_solved() indicates whether the model has been solved.
  • has_parsed_soln() indicates whether a parsed solution is
    available in the model.
  • display_soln() display the results of a solved model,
    including values, bounds and shadow prices for variables and
    linear constraints, values and shadow prices for nonlinear
    constraints, and individual cost components.

Bugs Fixed:

• Clear cached parameters after updating linear constraints or
  quadratic costs via opt_model.set_params().
• In miqps_mosek() the lower and upper bounds of binary variables
  got overwritten with 0 and 1, respectively, effectively relaxing
  any potentially tighter bounds provided as input.
• Fix false positive in have_feature_fsolve in case where the file
  is present, but without a valid license.

Other Changes:

• Update for compatibility with MATLAB R2023a (Optimization Toolbox
  9.5) and later, which removed x0 as a valid input to linprog.

• Update have_feature_ipopt() to recognize IPOPT MEX installations
  from Enrico Bertolazzi's mexIPOPT, which include MEX files
  that have been renamed to architecture-specific names along with an
  ipopt.m wrapper function to call the appropriate one.
  Thanks to Carlos Murillo-Sánchez.

  Note: While MP-Opt-Model no longer requires this, my
  recommendation is still to simply rename the MEX file to
  ipopt.<mexext>, with the appropriate architecture-specific
  extension, and delete the unnecessary ipopt.m entirely.

• Always skip price computation stage in miqps_<solver>() functions
  for pure (as opposed to mixed) integer problems.

• Add caching of aggregate output parameters in
  opt_model.params_var().

MP-Opt-Model 4.1

What's New in MP-Opt-Model 4.1

Released December 13, 2022

Below is a summary of the changes since version 4.0 of MP-Opt-Model. See
the CHANGES.md file for all the gory details. For release notes
for previous versions, see Appendix C of the MP-Opt-Model User's
Manual.

New Features

• Add support to qps_master() for calling qps_<my_solver>() by
  setting opt.alg to '<MY_SOLVER>' to allow for custom solvers.

Other Changes

• Update for compatibility with Artelys Knitro 13.1 and later.
• Add elapsed time in seconds to results of the solve() method of
  opt_model, returned in om.soln.output.et.
• Add runtime field to output argument of qps_glpk() and
  qps_mosek().

MP-Opt-Model 4.0

What's New in MP-Opt-Model 4.0

Released October 18, 2021

Below is a summary of the changes since version 3.0 of MP-Opt-Model. See
the CHANGES.md file for all the gory details. For release notes
for previous versions, see Appendix C of the MP-Opt-Model User's
Manual.

New Features

• Support for new class of problems -- parameterized nonlinear equations
  (PNE). Either create a model with only equality constraints (no
  inequalities or costs) and with number of variables equal to 1 more than
  number of constraints, or call pnes_master() directly. See Section 4.5
  of User's Manual for details.
  Thanks to Shrirang Abhyankar and Alexander Flueck for contributions to this
  feature, which is based on the continuation power flow code in MATPOWER 7.1.
  • Predictor/corrector numerical continuation method for tracing solution
    curves for PNE problems.
  • Plotting of solution curves.
  • User-defined event functions and callback functions.
  • Warm-start capabilities.
• Optional threshold for detecting failure of LEQ solve, by setting the
  leq_opt.thresh option. If the absolute value of any element of the
  solution vector exceeds the threshold, exitflag is set to 0, indicating
  failure.
• New functions:
  • pnes_master() provides unified interface for parameterized nonlinear
    equation (PNE) solvers.
  • pne_callback_default() collects PNE results and optionally plots
    solution curve.
  • pne_callback_nose() handles event signaling a nose point or limit
    has been reached.
  • pne_callback_target_lam() handles event signaling a target value
    of parameter λ has been reached.
  • pne_detect_events() detects events from event function values.
  • pne_detected_event() returns detected event details for events
    with a particular name.
  • pne_event_nose() detects the limit or nose point.
  • pne_event_target_lam() detects a target λ value.
  • pne_pfcn_arc_length() implements arc length parameterization.
  • pne_pfcn_natural() implements natural parameterization.
  • pne_pfcn_pseudo_arc_length() implements pseudo arc length
    parameterization.
  • pne_register_callbacks() registers callback functions.
  • pne_register_events() registers event functions.
  • mp_idx_manager/set_type_idx_map() method returns information about
    mapping of indices for a given set type back to the corresponding
    named (and possibly indexed) sets.
  • mpopt2pneopt() creates or modifies an options struct for
    pnes_master() from a MATPOWER options struct.

Bugs Fixed:

• Calling the problem_type() or is_mixed_integer() method on an empty
  model no longer causes a fatal error.

Other Changes

• Labels from the set_types property are now used as headers for
  opt_model/display() to simplify things facilitate use by sub-classes.
• Refactored describe_idx into a new method, set_type_idx_map, that
  returns in information in a programmatically usable form, and an updated
  describe_idx that calls the new method, then formats the results in
  the expected char array(s).

MP-Opt-Model 3.0

What's New in MP-Opt-Model 3.0

Released October 8, 2020

Below is a summary of the changes since version 2.1 of MP-Opt-Model. See
the CHANGES.md file for all the gory details. For release notes
for previous versions, see Appendix C of the MP-Opt-Model User's
Manual.

New Features

• Support for OSQP solver for LP and QP problems (https://osqp.org).
• Support for modifying parameters of an existing MP-Opt-Model object.
• Support for extracting specific named/indexed variables, costs, constraint
  values and shadow prices, etc. from a solved MP-Opt-Model object.
• Results of the solve() method saved to the soln field of the
  MP-Opt-Model object.
• Allow v0, vl, and vu inputs to opt_model/add_var() method, and
  l and u inputs to opt_model/add_lin_constraint() to be scalars
  that get expanded automatically to the appropriate vector dimension.
• New functions:
  • opt_model/set_params() method modifies parameters for a given named
    set of existing variables, costs, or constraints of an MP-Opt-Model
    object.
  • opt_model/get_soln() method extracts solved results for a given
    named set of variables, constraints or costs.
  • opt_model/parse_soln() method returns a complete set of solution
    vector and shadow price values for a solved model.
  • opt_model/eval_lin_constraint() method computes the constraint values
    for the full set or an individual named subset of linear constraints.
  • qps_osqp() provides standardized interface for using OSQP to
    solve LP/QP problems
  • osqp_options() initializes options for OSQP solver
  • osqpver() returns/displays version information for OSQP
  • ... plus 29 individual feature detection functions for
    have_feature(), see Table A-7 in the MP-Opt-Model User's Manual
    for details.

Bugs Fixed:

• Starting point supplied to solve() via opt.x0 is no longer ignored
  for nonlinear equations.
• Calling params_var() method with empty idx no longer results in
  fatal error.
• For opt_model, incorrect evaluation of constant term has been fixed for
  vector valued quadratic costs with constant term supplied as a vector.

Other Changes

• Simplified logic to determine whether a quadratic cost for an
  MP-Opt-Model object is vector vs. scalar valued. If the quadratic
  coefficient is supplied as a matrix, the cost is scalar varied,
  otherwise it is vector valued.
• Deprecated have_fcn() and make it a simple wrapper around the new
  modular and extensible have_feature(), which has now been moved to
  MP-Test (https://github.com/MATPOWER/mptest).

MP-Opt-Model 2.1

What's New in MP-Opt-Model 2.1

Released August 25, 2020

Below is a summary of the changes since version 2.0 of MP-Opt-Model. See
the CHANGES.md file for all the gory details. For release notes
for previous versions, see Appendix C of the MP-Opt-Model User's
Manual.

New Features

• Fast-decoupled Newton's and Gauss-Seidel solvers for nonlinear equations.
• New linear equation ('LEQ') problem type for models with equal
  number of variables and linear equality constraints, no costs, and no
  inequality or nonlinear equality constraints. Solved via mplinsolve().
• The solve() method of opt_model can now automatically handle mixed
  systems of equations, with both linear and nonlinear equality constraints.
• New core nonlinear equation solver function with arbitrary, user-defined
  update function, used to implement Gauss-Seidel and Newton solvers.
• New functions:
  • nleqs_fd_newton() solves a nonlinear set of equations via a
    fast-decoupled Newton's method.
  • nleqs_gauss_seidel() solves a nonlinear set of equations via a
    Gauss-Seidel method.
  • nleqs_core() implements core nonlinear equation solver with
    arbitrary update function.

Incompatible Changes

• In output return value from nleqs_newton(), changed the normF
  field of output.hist to normf, for consistency in using lowercase
  f everywhere.

MP-Opt-Model 2.0

What's New in MP-Opt-Model 2.0

Released July 8, 2020

Below is a summary of the changes since version 1.0 of MP-Opt-Model. See
the CHANGES.md file for all the gory details. For release notes
for previous versions, see Appendix C of the MP-Opt-Model User's
Manual.

New Features

• Add new 'fsolve' tag to have_fcn() to check for availability of
  fsolve() function.
• Add nleqs_master() function as unified interface for solving
  nonlinear equations, including implementations for fsolve and
  Newton's method in functions nleqs_fsolve() and nleqs_newton(),
  respectively.
• Add support for nonlinear equations (NLEQ) to opt_model. For
  problems with only nonlinear equality constraints and no costs,
  the problem_type() method returns 'NLEQ' and the solve()
  method calls nleqs_master() to solve the problem.
• New functions:
  • mpopt2nleqopt() creates or modifies an options struct for
    nleqs_master() from a MATPOWER options struct.
  • nleqs_fsolve() provides implementation of unified nonlinear
    equation solver interface for fsolve.
  • nleqs_master() provides a single wrapper function for calling
    any of MP-Opt-Model's nonlinear equation solvers.
  • nleqs_newton() provides implementation of Newton's method solver
    with a unified nonlinear equation solver interface.
  • opt_model/params_nln_constraint() method returns parameters for
    a named (and optionally indexed) set of nonlinear constraints.
  • opt_model/params_nln_cost() method returns parameters for a
    named (and optionally indexed) set of general nonlinear costs.

Other Changes

• Add to eval_nln_constraint() method the ability to compute constraints
  for a single named set.
• Skip evaluation of gradient if eval_nln_constraint() is called with
  a single output argument.
• Remove redundant MIPS tests from test_mp_opt_model.m.
• Add tests for solving LP/QP, MILP/MIQP, NLP and NLEQ problems via
  opt_model/solve().
• Add Table 6-1 of valid have_fcn() input tags to User's Manual.

MP-Opt-Model 1.0

What's New in MP-Opt-Model 1.0

Released May 8, 2020

Below is a summary of the changes since version 0.8 of MP-Opt-Model. See the CHANGES.md file for all the gory details. For release notes for previous versions, see Appendix C of the MP-Opt-Model User's
Manual.

New Documentation:

• Add the MP-Opt-Model User's Manual with LaTeX source code included in docs/src.

Other Improvements:

• Refactor opt_model class to inherit from new abstract base class mp_idx_manager which can be used to manage the indexing of other sets of parameters, etc. in other contexts.