v2.2.2 pip support and parallel updated help

This release updates the help message for hypermapper --help to be more helpful for users.

==Changes in pip support and parallel MSLS release (v2.2)

This release adds pip support to HyperMapper. HyperMapper can now be installed with
pip install hypermapper

This release also contains the following pip-related changes:

For pip-installations, the HyperMapper main script (formerly called via python scripts/hypermapper.py can now be called by the command hypermapper from the command line.
We also added the following command line hooks for pip installations:
    scripts/compute_pareto.py can now be be called by hm-compute-pareto.
    scripts/plot_pareto.py can now be called by hm-plot-pareto.
    scripts/plot_hvi.py can now be called by hm-plot-hvi.
    The quick start scenario in example_scenarios/quick_start/branin.py can now be called from the command line by hm-quickstart.
    scripts/plot_optimization_results.py can now be called by hm-plot-optimization-results.
The pip installation does not contain the example scenarios, apart from the quick start scenario. However, if you clone the repository to get the examples, make sure to call them as a module (see below). This way, you don’t need to set any environment variables with the pip installation.
The pip installation does not require any environment variables.

We recommend executing the example scenarios by executing them as a module, e.g., by python -m example_scenarios.clients.python.client-server_chakong_haimes.

It is still possible to call scripts/hypermapper.py as well as the other scripts in the scripts folder as before. For this to work, you need to
either

call the scripts from the HyperMapper root directory,
or to set HYPERMAPPER_HOME to your HyperMapper root directory.

We recommend deleting the reference to HYPERMAPPER_HOME/scripts from PYTHONPATH although an existing reference should not break anything with this release (except for the rare case where you have set this reference but not HYPERMAPPER_HOME).

We changed the directory structure. The legacy scripts are in the scripts folder. These import everything from the new hypermapper directory. The scripts directory is not part of the pip installation. For pip installations, the scripts can be executed using the command line hooks (see above).

This release also contains optimization and better parallelism support, which makes HyperMapper significantly faster.

At last, this release also adds support for noiseless GP models, which can be used via the json configuration file.

v2.2.1 pip support and parallel MSLS fixes

This release fixes a couple of small bugs in the pip support and parallel MSLS release. Functionality remains unchanged.

==Changes in pip support and parallel MSLS release (v2.2)

This release adds pip support to HyperMapper. HyperMapper can now be installed with
pip install hypermapper

This release also contains the following pip-related changes:

For pip-installations, the HyperMapper main script (formerly called via python scripts/hypermapper.py can now be called by the command hypermapper from the command line.
We also added the following command line hooks for pip installations:
    scripts/compute_pareto.py can now be be called by hm-compute-pareto.
    scripts/plot_pareto.py can now be called by hm-plot-pareto.
    scripts/plot_hvi.py can now be called by hm-plot-hvi.
    The quick start scenario in example_scenarios/quick_start/branin.py can now be called from the command line by hm-quickstart.
    scripts/plot_optimization_results.py can now be called by hm-plot-optimization-results.
The pip installation does not contain the example scenarios, apart from the quick start scenario. However, if you clone the repository to get the examples, make sure to call them as a module (see below). This way, you don’t need to set any environment variables with the pip installation.
The pip installation does not require any environment variables.

We recommend executing the example scenarios by executing them as a module, e.g., by python -m example_scenarios.clients.python.client-server_chakong_haimes.

It is still possible to call scripts/hypermapper.py as well as the other scripts in the scripts folder as before. For this to work, you need to
either

call the scripts from the HyperMapper root directory,
or to set HYPERMAPPER_HOME to your HyperMapper root directory.

We recommend deleting the reference to HYPERMAPPER_HOME/scripts from PYTHONPATH although an existing reference should not break anything with this release (except for the rare case where you have set this reference but not HYPERMAPPER_HOME).

We changed the directory structure. The legacy scripts are in the scripts folder. These import everything from the new hypermapper directory. The scripts directory is not part of the pip installation. For pip installations, the scripts can be executed using the command line hooks (see above).

This release also contains optimization and better parallelism support, which makes HyperMapper significantly faster.

At last, this release also adds support for noiseless GP models, which can be used via the json configuration file.

v2.2 pip support and parallel MSLS

This release adds pip support to HyperMapper. HyperMapper can now be installed with
pip install hypermapper

This release also contains the following pip-related changes:

• For pip-installations, the HyperMapper main script (formerly called via python scripts/hypermapper.py can now be called by the command hypermapper from the command line.
• We also added the following command line hooks for pip installations:
  • scripts/compute_pareto.py can now be be called by hm-compute-pareto.
  • scripts/plot_pareto.py can now be called by hm-plot-pareto.
  • scripts/plot_hvi.py can now be called by hm-plot-hvi.
  • The quick start scenario in example_scenarios/quick_start/branin.py can now be called from the command line by hm-quickstart.
  • scripts/plot_optimization_results.py can now be called by hm-plot-optimization-results.
• The pip installation does not contain the example scenarios, apart from the quick start scenario. However, if you clone the repository to get the examples, make sure to call them as a module (see below). This way, you don’t need to set any environment variables with the pip installation.
• The pip installation does not require any environment variables.

We recommend executing the example scenarios by executing them as a module, e.g., by python -m example_scenarios.clients.python.client-server_chakong_haimes.

It is still possible to call scripts/hypermapper.py as well as the other scripts in the scripts folder as before. For this to work, you need to
either

• call the scripts from the HyperMapper root directory,
• or to set HYPERMAPPER_HOME to your HyperMapper root directory.

We recommend deleting the reference to HYPERMAPPER_HOME/scripts from PYTHONPATH although an existing reference should not break anything with this release (except for the rare case where you have set this reference but not HYPERMAPPER_HOME).

We changed the directory structure. The legacy scripts are in the scripts folder. These import everything from the new hypermapper directory. The scripts directory is not part of the pip installation. For pip installations, the scripts can be executed using the command line hooks (see above).

This release also contains optimization and better parallelism support, which makes HyperMapper significantly faster.

At last, this release also adds support for noiseless GP models, which can be used via the json configuration file.

v2.1.1 Evolutionary algorithm, new client examples, and Batch BO

This release adds a few functionalities to HyperMapper:

• A new evolutionary algorithm that can be used for optimizing black-box functions
• A new Batch Bayesian Optimization algorithm that allows HyperMapper to request multiple black-box functions evaluations in parallel
• New and updated client-server mode examples
• A new plotting script for mono-objective optimization
• Several minor code optimizations

HyperMapper v2.1: Prior-guided Optimization

This release adds a new Prior-guided Optimization method to HyperMapper, an optimization approach that allows users to inject their knowledge into the optimization process in the form of priors about which parts of the input space will yield the best performance.

This release also adds a new resume optimization feature to HyperMapper, which allows users to resume a previous optimization run of HyperMapper.

HyperMapper v2.0

This release brings several updates to HyperMapper, namely:

• A new optimization approach based on random scalarizations, which supports optimization for any number of objectives
• An updated Random Forest model with better performance
• A new Gaussian Process model
• Support for constrained Bayesian Optimization
• A new default mode of operation, that allows HyperMapper to call the black-box function directly during optimization
• A new optimization approach based on Multi-start Local Searches that also supports multi-objective optimization
• Multiple fixes and performance improvements
• Several new optimization examples