https://github.com/coin-or/rbfopt - Stable v4.2.6

Changed:

• Configuration for the documentation build system.

- Python
Published by gnannicini about 3 years ago

https://github.com/coin-or/rbfopt - Stable v4.2.5

Changed:

• Improved efficiency of parallel initialization: now all function evaluations in the initialization procedure (as well as restarts) are handled asynchronously, and optimization can start even if the initialization phase is not complete.

Added:

• Added parameter initsampleincrease_parallel to control how many additional samples should be generated in the initialization phase when running in parallel mode. This leads to increased efficiency of CPU usage.

- Python
Published by gnannicini over 3 years ago

https://github.com/coin-or/rbfopt - Stable v4.2.3

Changed:

• Fixed bug in parallel optimization with a noisy objective function, that could lead to a crash in some situations (generally triggered by running on many threads).
• Slightly more efficient code to update some data structures in rbfopt_algorithm.

- Python
Published by gnannicini over 4 years ago

https://github.com/coin-or/rbfopt - Stable v4.2.2

Changed:

• Fixed bug in the genetic algorithm when optimizing functions with categorical variables at the end of the list of variables.
• Updated documentation.

- Python
Published by gnannicini over 5 years ago

https://github.com/coin-or/rbfopt - Stable v4.2.0

Added:

• Support for categorical variables (variable type 'C', in addition to 'R' for real and 'I' for integer). These variables are treated differently from integer variables, but they still need to have lower and upper bounds; e.g., a categorical variable with bounds [0, 2] can take on 3 possible values: 0, 1, 2.

Changed:

• Renamed parameter names starting with 'tr' to 'ref', for better consistency with the literature.

- Python
Published by gnannicini about 6 years ago

https://github.com/coin-or/rbfopt - stable v4.1.2

Added:

• Additional test functions and unit tests.

Changed:

• Better pickling with dill, if the package is installed.

Bugfixes:

• Fixed bug with uninitialized bestglobalrbf_list when the linear system is singular.

- Python
Published by gnannicini over 6 years ago

https://github.com/coin-or/rbfopt - stable v4.1.1

Bugfixes:

• Fixed bug with sample_size not being cast to integer and raising an error.

- Python
Published by gnannicini over 7 years ago

https://github.com/coin-or/rbfopt - stable v4.1.0

Added:

• Finer control of the optimization process, via several options: initincludemidpoint, initsamplefraction, dolocalsearch.

Changed:

• More efficient initialization phase, requiring fewer points.

- Python
Published by gnannicini over 7 years ago

https://github.com/coin-or/rbfopt - stable v4.0.3

Added:

• More ways for the user to control the initial points and optimization process.
• Ability to control debug messages from solvers.

Changed:

• Format of the output log is now more compact.
• Improved management of failed/discarded iterations.

Bugfixes:

• Fixed bug in parallel optimizer when evaluating RBF model after restarts.

- Python
Published by gnannicini almost 8 years ago

https://github.com/coin-or/rbfopt - stable v4.0.2

Added:

• Detailed user manual.
• Improved testing tools.

Changed:

• Smaller formulation for the noisy interpolation models.
• Improved stability of the restoration step.

Bugfixes:

• Fixed bug in node deletion in rbfopt_algorithm.
• Fixed infinite loop when testing multiprocessing on Windows.
• Fixed unexpected exit in restoration phase.

- Python
Published by gnannicini about 8 years ago

https://github.com/coin-or/rbfopt - stable v4.0.1

Bugfixes:

• Updated function names in rbfoptuserblack_box to match base abstract class.

- Python
Published by gnannicini over 8 years ago

https://github.com/coin-or/rbfopt - stable v4.0.0

Added:

• Support for Gaussian radial basis function.
• Class to construct RbfoptBlackBox from user data (rbfoptuserblack_box).
• The shape of the multiquadric and Gaussian RBF can now be set at runtime using the rbfshapeparameter option in RbfoptSettings.
• Automatic selection of the shape parameter.
• Option maxcrossvalidations to limit the total number of cross validations performed. ### Changed:
• Improved numerical stability for nonlinear models in Gutmann's algorithm.
• Cross validation now uses Gaussian RBFs and different values of the shape parameter.
• BlackBox renamed to RbfoptBlackBox.
• RbfoptBlackBox interface now requires returning var_type instead of a list of integer vars.
• RbfoptBlackBox.evaluatefast is now called evaluatenoisy, and requires returning an estimate of the error.
• Documentation and examples are now distributed within the rbfopt module.
• README updated with new installation instructions and minimal examples. ### Removed:
• Module rbfopt_config is no longer necessary.

- Python
Published by gnannicini almost 9 years ago

https://github.com/coin-or/rbfopt - stable v3.0.2

Added:

• Model improvement search for refinement phase. ### Changed:
• Refinement phase now happens regardless of finding an improved solution.
• Pyomo models changed for numerical stability with default Ipopt and Bonmin as downloaded from AMPL's website.

- Python
Published by gnannicini almost 9 years ago

https://github.com/coin-or/rbfopt - stable v3.0.1

Changed:

• Faster and improved refinement phase.
• Several default settings have been changed to exploit the refinement phase further.
• Fixed bug in parallel optimizer.

- Python
Published by gnannicini about 9 years ago

https://github.com/coin-or/rbfopt - stable v3.0.0

Major update with several improvements. Not backward compatible.

Added:

• Refinement phase for improved local search: see option refinement_frequency and new options that start with tr_.
• Support for Python 3.

Changed:

• OptAlgorithm is now called RbfoptAlgoritm.
• RbfSettings is now called RbfoptSettings.
• Automatic model selection is now active by default, using fast linear algebra routines.
• Rbf functions in module rbfopt_utils now use Numpy, triggering several code optimization changes and major improvement in iteration speed.
• Several default settings have been changed.

Removed:

• Module rbfoptmodelselection is no longer necessary.
• Option modelselectionsolver is no longer necessary.
• Option maxstalledobjfunimpr is subsumed by eps_impr.
• Cython modules and setup.py have been removed .

- Python
Published by gnannicini about 9 years ago

https://github.com/coin-or/rbfopt - stable v2.1.0

New features:

• Faster iteration times, performing heave calculations with NumPy.
• Support for compiling crucial modules with Cython, further increasing iteration speed.

This comes with the usual fixes and improvements, and some changes to default parameter values.

- Python
Published by gnannicini over 9 years ago

https://github.com/coin-or/rbfopt - stable v2.0.0

New features: - More search options, including vastly improved default algorithms. - Parallel version of the optimization engine, allowing asynchronous parallel evaluation of the black-box function. - Greater numerical stability. - Improved command-line interface allowing easier execution on user-defined problems.

This major version is not backward compatible, due to changes in the interface to black box problems that were made necessary to handle parallelization. Please see the file rbfoptblackbox_example.py to see how to implement a black-box function.

- Python
Published by gnannicini almost 10 years ago

https://github.com/coin-or/rbfopt - stable v1.2.0

New features: - A new, powerful optimization algorithms: Metric SRSM (based on work from Regis & Shoemaker). - Improved stability thanks to extensive bugfixing (mostly for strange parameter combinations).

- Python
Published by gnannicini over 10 years ago

https://github.com/coin-or/rbfopt - stable v1.1.0

Features: - Uses Pyomo 4. - Support CPLEX and Clp for faster automated model selection. - Several improvements and fixes. - Improved parameter documentation and help.

WARNING: You will have to update to Pyomo 4, which does not work with previous Pyomo 3 code. Furthermore, parameter names for the command line interface have been changed to match those in the code. This will be the approach followed from now on. The rest of the API is backward compatible and previous code that uses the API should work after updating to Pyomo 4.

- Python
Published by gnannicini almost 11 years ago

https://github.com/coin-or/rbfopt - stable v1.0.2

Supports unconstrained optimization

- Python
Published by gnannicini about 11 years ago