NOMAD 4.4.0

NOMAD 4 is available as open-source code, under the LGPL license. Compilation on MacOS, Windows and Linux is done using cmake.
Binary packages are also available.

We are looking forward for user feedback.
For more information, reach out to [email protected].

The user guide is available here.

We recommend to download the complete NOMAD source files and examples and build the project for your platform. The user guide describes the complete steps for building.

For users who do not follow the previous recommendation, a compact version with binaries (zipped) are available for Windows, Mac-OSX and Linux Ubuntu in the Assets section below. Please note, that you will need to do manual modifications to execute the binaries (see the end Note in the README file).

New in Release 4.4:
DiscoMads algorithm to reveal and escape some regions with hidden constraints or containing discontinuities.
Callbacks functions for controlling runs.
PyNomad can be installed from PyPi.org

NOMAD 4.3.1

NOMAD 4 is available as open-source code, under the LGPL license. Compilation on MacOS, Windows and Linux is done using cmake.
Binary packages are also available.

We are looking forward for user feedback.
For more information, reach out to [email protected].

The user guide is available here.

We recommend to download the complete NOMAD source files and examples and build the project for your platform. The user guide describes the complete steps for building.

For users who do not follow the previous recommendation, a compact version with binaries (zipped) are available for Windows, Mac-OSX and Linux Ubuntu in the Assets section below. Please note, that you will need to do manual modifications to execute the binaries (see the end Note in the README file).

New in Release 4.3:
DMultiMads algorithm for multiobjective optimization problems.
A Template algorithm to help users implement their own method in Nomad.
Static surrogate can be used in library mode.
Enable control the display precision.

NOMAD 4.3.0

NOMAD 4 is available as open-source code, under the LGPL license. Compilation on MacOS, Windows and Linux is done using cmake.
Binary packages are also available.

We are looking forward for user feedback.
For more information, reach out to [email protected].

The user guide is available here.

We recommend to download the complete NOMAD source files and examples and build the project for your platform. The user guide describes the complete steps for building.

For users who do not follow the previous recommendation, a compact version with binaries (zipped) are available for Windows, Mac-OSX and Linux Ubuntu in the Assets section below. Please note, that you will need to do manual modifications to execute the binaries (see the end Note in the README file).

New in Release 4.3:
DMultiMads algorithm for multiobjective optimization problems.
A Template algorithm to help users implement their own method in Nomad.
Static surrogate can be used in library mode.
Enable control the display precision.

NOMAD 4.2.0

NOMAD 4 is available as open-source code, under the LGPL license. Compilation on MacOS, Windows and Linux is done using cmake.
Binary packages are also available.

We are looking forward for user feedback.
For more information, reach out to [email protected].

The user guide is available here.

We recommend to download the complete NOMAD source files and examples and build the project for your platform. The user guide describes the complete steps for building.

For users who do not follow the previous recommendation, a compact version with binaries (zipped) are available for Windows, Mac-OSX and Linux Ubuntu in the Assets section below. Please note, that you may need to change the permissions to be able to execute the binaries.

New in Release 4.2:

• Poll direction ORTHO N+1 QUAD is now supported for parameters DIRECTION_TYPE.
• Default evaluation points sorting uses quadratic model.
• PSD-Mads has been implemented.
• Java interface is supported using Swig.
• Building PyNomad interface is supported for all versions using cmake.
• Building Matlab interface is done with cmake.

NOMAD 4.1.0

NOMAD 4 is available as open-source code, under the LGPL license. Compilation on MacOS, Windows and Linux is done using cmake.

We are looking forward for user feedback.
For more information, reach out to [email protected].

The doc is available here https://nomad-4-user-guide.readthedocs.io/en/v4.1.0/

New in Release 4.1:

• It is possible to use a static surrogate executable to sort points before evaluating them with the blackbox. See parameter EVAL_QUEUE_SORT with value SURROGATE.
• It is also possible to do a full optimization using only the surrogate executable instead of the blackbox executable. See parameter EVAL_SURROGATE_OPTIMIZATION.
• Poll direction ORTHO N+1 NEG is now supported for parameters DIRECTION_TYPE and DIRECTION_TYPE_SECONDARY_POLL. It is also now possible to define several direction types for these parameters.
• Variable Neighborhood Search, using Mads for sub-optimization, is now supported. See parameter VNS_MADS_SEARCH.
• NOMAD can now be compiled on Windows. Follow the instructions in the README. PyNomad interface is not supported for this version.

NOMAD 4.0.2

NOMAD 4 official release.

NOMAD 4 is available as open-source code, under the LGPL license. Compilation on MacOS and Linux is done using cmake. The code has not been compiled on Windows.

We are looking forward for user feedback.
For more information, reach out to [email protected].

The doc is available here https://nomad-4-user-guide.readthedocs.io/en/v.4.0.2/

New in Release 4.0:

• Secondary Poll center
• Speculative Search as in NOMAD 3
• Sort on direction of last success
• Latin hypercube may be used when X0 is not provided
• Updated parameter names
• User guide

Features already available in NOMAD 4, Beta 2:

• Quadratic Model Search (QUAD_MODEL_SEARCH)
• Groups of variables (VARIABLE_GROUP)
• Direction types (DIRECTION_TYPE)
• Sort evaluation queue before evaluating points; randomize point evaluation
• PSD-Mads
• Additional statistics, HISTORY_FILE, SOLUTION_FILE
• Compilation using CMake
• Parameter USE_CACHE

Features already available in NOMAD 4, Beta 1:

• Parameter syntax for parameter file is the same as in NOMAD 3
• Search implementations: Nelder Mead, Speculative, Latin Hypercube, SgtelibModel
• Usage of library Sgtelib
• Management of PB (progressive barrier) and EB (extreme barrier) constraints
• Python interface
• Callbacks
• Block evaluations
• Evaluations are done in parallel, using OpenMP
• Hot restart: See parameter HOT_RESTART_ON_USER_INTERRUPT

NOMAD 4.0.1

Please use NOMAD 4.0.2 or later

NOMAD 4.0.0

Please use NOMAD 4.0.2 or later

NOMAD 4.0.0 Beta 2

NOMAD 4 Beta 2 has numerous improvements since Beta 1 (December 2019).

NOMAD 4 is available as open-source code, under the LGPL license. Makefiles are present for compilation on MacOS and Linux. The code has not been compiled on Windows.

We are looking forward for user feedback.
For more information, reach out to [email protected].

Overview:

New in Beta 2:

• Quadratic Model Search (QUAD_MODEL_SEARCH)
• Groups of variables (VARIABLE_GROUP)
• Direction types (DIRECTION_TYPE)
• Sort evaluation queue before evaluating points; randomize point evaluation
• PSD-Mads
• Additional statistics, HISTORY_FILE, SOLUTION_FILE
• Compilation using CMake
• Parameter USE_CACHE

Features already available in NOMAD 4, Beta 1:

• Parameter syntax for parameter file is the same as in NOMAD 3
• Search implementations: Nelder Mead, Speculative, Latin Hypercube, SgtelibModel
• Usage of library Sgtelib
• Management of PB (progressive barrier) and EB (extreme barrier) constraints
• Python interface
• Callbacks
• Block evaluations
• Evaluations are done in parallel, using OpenMP
• Hot restart: See parameter HOT_RESTART_ON_USER_INTERRUPT

v.4.0.0-beta.2

Merge pull request #9 from bbopt/develop

Beta 2