Skip to content

Latest commit

 

History

History
277 lines (229 loc) · 11.1 KB

CHANGELOG.md

File metadata and controls

277 lines (229 loc) · 11.1 KB
Raw

Alpine.jl Change Log

v0.5.6

  • Minor Readme updates (#235)
  • Implemented RawStatusString for solution summary (#238)
  • Dependency update for docs
  • Bug fix in LB exceeding UB outside relative tolerance bounds

v0.5.5

  • Removed use of e.g. for JuMP documentation in README (#230)
  • Dropped support for integer variable MINLPs in examples (#233)
  • Updated README in examples folder to clarify minlp_solver dependency (#232)

v0.5.4

  • Added random_QCQPs folder with python generator files for all the instances appearing in this paper: https://arxiv.org/pdf/2301.00306.pdf
  • Minor tweaks in docs and main README file

v0.5.3

  • Removed compatability on Pkg #225
  • Added raw NL expression interface example for a dense multilinear problem

v0.5.2

  • Fixed issue with @NLconstraint created from a quadratic @expression (#221)
  • Fixed error message in nlexpr parsing of constraints with non-integer exponents (#223)
  • Clean up in expr_is_fractional_exponent (related to #223)
  • Added dependency on Pkg package
  • Added unit tests for testing multilinear binary products (hmittelman)

v0.5.1

  • Re-org and clean-up in partition additions within bounding MIP model
  • Updated to cheaper AffExpr evaluation within linking constraints
  • Minor update in gap evaluataion after OBBT termination with fixed iterations
  • Added solver option use_start_as_incumbent to use warm start value as also an intial incumbent solution in presolve without invoking a local solver. (#218)
  • Added unit tests for testing warm start point

v0.5.0

  • New feature: Linking constraints for multilinear terms with uniform and adaptive partitions (significant speed up in run times for multilinear problems: http://www.optimization-online.org/DB_HTML/2022/07/8974.html) (@jongeunkim)
  • Added dependency on Combinatorics package for linking constraints
  • Added unit tests for linking constraints feature
  • Changed Cbc test dependency to HiGHS solver
  • Dropped a few redundant unit tests
  • Dropped support for (slower) piecewise McCormick formulations
  • Changed disc_ratio to partition_scaling_factor
  • Added JuliaFormatter.toml and formatting workfow

v0.4.3

  • Dropped support for checking solver availability. User can choose to use any MILP, MINLP and/or NLP solver with Alpine
  • GRB_ENV now works with Gurobi solver to avoid printing License info multiple times during iterations
  • Updates tests to reflect above changes
  • Default convhull_warmstart set to true for a MIP solver in bounding-MIP iterations

v0.4.2

  • Fix support for JuMP v1.2 (@odow)

v0.4.1

  • Added MOI attribute for tightened bounds post-OBBT (@blegat)

v0.4.0

  • Fixed OBBT convergence issue due to average bound reduction
  • Added new termination criteria to OBBT: max_reduction, max_width, rel_gap at every iteration of OBBT
  • Fixed the issue to tighten only original nonlinear variables (aux vars were counted before)
  • presolve_bt_bound_tol changed to 1E-4 (numerically stable)
  • presolve_bt_obj_bound_tol added to avoid infeasibility in OBBT iterations when bounds are tights
  • Parameter tuning (tolerances) for stable OBBT convergence
  • Fixed bug in best_rel_gap evaluation
  • Bug fixed to check any user-defined operators, and disable hessians based on availabe features. Unit test added for this
  • Now, Alpine terminates (w/o invoking global_solve) when presolve finds the global optimum due to rel_gap eval at every OBBT iteration. Unit test added for this.
  • Post-OBBT optimality gap (presolve_best_rel_gap) user-option added to be able to extract best opt gap after OBBT when Alpine terminates
  • apply_partitioning feature added to be able to terminate Alpine without applying the MIP-based paritioning algorithm (mainly to run only OBBT)
  • Revised unit test (in test_algorithm) for reduced run times

v0.3.0

  • JuMP v1.0+ in examples
  • Pavito v0.3.5 support
  • Dropped support for redundant run_bounding_iteration function
  • Clean-up and re-org of optimizer structs and MOI_wrapper functions
  • Bug fix in relax_integrality under presolve
  • Added Rosenbrock function
  • Added user-definable presolve_bt_improv_tol for variable range reductions in OBBT
  • Minor update on finite values for obj-bound in create_obbt_model
  • NLP solution rounding issue resolved (now compatible with default tol, 1e-6), to avoid the variable solution outside discretization (#190)
  • circleN instance updated to circle_MINLPLib to eliminate Pavito's mixed-integer cycling in unit tests (#190)
  • Clean-up in printing variable solutions (variable_values)

v0.2.10

  • Add support for the HiGHS solver

v0.2.9

  • Update to JuMP v1.0

v0.2.8

  • Update to JuMP v0.22, v0.23
  • Update to MOI v1.0
  • Simplify test dependencies

v0.2.7

  • Support for module shortcut Alp in function calls
  • Solver options clean up in examples folder
  • Dropped support for unsued functions: _bound_sense update_boundstop_options amp_post_inequalities_int amp_pick_ratevec amp_collect_tight_regions
  • Improved code coverage
  • Docs cleanup

v0.2.6

  • Bug fix in algorithm.jl to handle Gurobi as the MIP solver with updated MOI
  • Deactivated redundant functions collect_lb_pool, merge_solution_pool and track_new_partition_idx
  • Moved examples out of test folder
  • Updated run_examples to handle Gurobi as the MIP solver
  • Removed dependency on the Distributed package
  • Improved code coverage
  • Typos fixed in docs

v0.2.5

  • Cleaning up of solver logs
  • Time correction for local solve
  • Removed unused Alpine.Optimizer options in solver.jl

v0.2.4

  • Added unit tests for utility functions
  • Removed un-used functions from utility functions which aren't part of Alpine's algorithm
  • Dropped support for heuristic rounding algorithm (heu_basic_rounding) - minlp_solver input is a must
  • Dropped support for adjusting branching priority in MIP solvers (adjust_branch_priority)

v0.2.3

  • Migrating from Travis to Github-actions
  • Major documentation clean-up
  • Dependency packages clean-up
  • runtests.jl clean-up
  • Solver options update: presolve_max_iter => presolve_bt_max_iter and others
  • Added test/examples/run_example.jl

v0.2.2

  • README clean up

Closed issues:

  • Constraints are not supported and cannot be bridged into supported constrained variables and constraints (#164)

v0.2.1

  • Logging clean up
  • Removal of redundant tests

Closed issues:

  • Juniper's ALMOST_LOCALLY SOLVED error (#157)
  • unncessary eval usages (#153)
  • Logging issue with constants in objective (#159)

v0.2.0

  • MOI wrapper and JuMP v0.21 support
  • expression.jl clean up
  • eval() removal where unnecessary (#153)
  • Drop support for all trigonometric related functions (both in src and testing)
  • Drop support of expression parsing for generic integer variables in MINLPs (both in src and testing)

Closed issues:

  • arc consistency partitioning pruning scheme as an example extension for POD (#37)
  • Generate convex relaxations for nonconvex MINLPs (#86)
  • Error in convexification of monomials (#104)
  • On variables chosen for partitioning (#121)
  • Travis and Documeter Maintenance (#144)
  • Compilation error (#146)
  • bound propagation buggy (#147)
  • Reenable test: Validation Test || PBT-AMP-TMC || basic solve || examples/nlp3.jl (#150)
  • Reenable test: Validation Test || AMP-CONV-FACET || basic solve || examples/nlp1.jl (#151)

v0.1.16

  • Fixed CBC version to make it compatible
  • Removed support for "Mini" Formulation for convex hulls
  • Updated tests and removed a few

v0.1.15

  • Bug fixes for issues 134 and 135 (fractional exponents, constants within integer exponents)
  • Added function 'expr_isolate_const' to nlexpr.jl
  • Meaningful optimality gaps for zero upper bound values (issue 103)
  • Typos in comments
  • Dropped support for colorful_alpine
  • Ole Kroger's updates to docs/make.jl

v0.1.10

  • Multiple bug fixes in expression parsing
  • Added support for feasibility problems with constant objectives
  • Typos in comments

v0.1.9

  • Drop support for Julia < v1.0.
  • Move to the new Registrator system.
  • The last supported version for Julia v0.6 is v0.1.8
  • The last supported version for Julia v0.7 is v0.1.7 (the bugs fixed in v0.1.8 is not ported to Julia v0.7)

v0.1.8

  • Bug fixes in solver recognition
  • Updated Issue template and Readme.
  • Rounding bug fixed for binaries.
  • This version is only available for Julia v0.6

v0.1.7

  • Bug fixes for bounds on lifted variables representing nonlinear terms
  • Inf bound warning issue fixed

v0.1.6

  • Bug fix in operators.jl for accounting all extreme points in the bound evaluation

v0.1.5

  • Unit exponent bug-fix
  • Meaningful error messages for unsupported functions
  • Pinned tests to JuMP 0.18.5
  • Cleaner & meaningful logs

v0.1.3

  • Finite large bounds on unbounded variables

Staged

  • typo fix
  • set Big-M values for unbounded variables

v0.1.2

  • Bug fix in bound tightening tolerance corner case

v0.1.1

  • Bug fix in Cplex solver recognition

v0.1.0

  • Alpine.jl first release

Merged Features

  • partition_scaling_factor selection algorithm
  • logarithm embedding formulation
  • bounding model warm starting and no-good-cuts using solution pool
  • disc_vars selection algorithm based on weighted minimum vertex cover
  • more support for binary problem
  • rounding heuristic and multi-start heuristic for feasible solutions
  • recursive linear lifting
  • some attempts to reason bounds for variables with infinite bounds

New Operators

  • BINPROD: product of binary vars
  • BINLIN: product of a binary var and continuous var

Data Structure

  • Some refactoring to shorten the attribute names
  • Eliminated unnecessary/obsolete user options
  • Strengthen the checks to problems

Code Structure

  • Alpine.jl: "header" file
  • solver.jl: data structure and problem loading
  • algorithm.jl: high-level algorithms and major utilities (presolve(), local_solve(), bounding_solve())
  • amp.jl: algorithms for bounding formulation constructions
  • presolve.jl: algorithms for presolves
  • heuristics.jl: algorithms for heuristics
  • bounds.jl: algorithms and utilities for handling bounds including bounds propagation and reasoning with lifting
  • mpb2moi.jl: preparation for migration to MOI interface, a collection of all solver & modeling interface warpping functions
  • multi.jl: high-level piecewise convex relaxation formulation construction and utilities for multilinear/monomial terms
  • tmc.jl: old McCormick-based methods
  • const.jl: (experimental) tries to categorize operators for cleaner and safer operations
  • embedding.jl: logarithm embedding formulations and utilities
  • nlexpr.jl: algorithms for expression handling
  • operators.jl: operator detection algorithm and associated utilities for bounds, discretization variable selection, etc.
  • log.jl: logging utilities
  • utility.jl: general utility functions

Debugs

  • Fixed multiple bugs in recursive expression handling
  • Fixed duplicates in constructing convex relaxations
  • Fixed bound reasoning steps for complex terms
  • Fixed many other things

Tests & Coverage

  • Most new features are covered (embedding, partition_scaling_factor algorithm, etc.)
  • Stronger expression parsing tests with new operators covered
  • Coverage should be more than ~90%