The **H**ighly **O**ptimized **P**olytope **S**ampling toolbox is an open-source C++17 library for efficient and scalable MCMC algorithms for sampling convex-constrained spaces possibly equipped with arbitrary target functions.

For details and benchmarks see the application note https://doi.org/10.1093/bioinformatics/btaa872. Test data from the application note is downloadable at https://doi.org/10.26165/JUELICH-DATA/YXLFKJ.

Documentation

Documentation, interactive demos and further resources can be found at https://modsim.github.io/hops/.

Cloning from Github

HOPS contains git submodules that point to third-party libraries. For this reason, HOPS should be fetched recursively:

git clone git@github.com:modsim/hops.git --recursive

HOPS Logo

Installation

HOPS uses CMake as build system.
See the Dockerfile for a demonstration on installing HOPS and its dependencies on Ubuntu 20.4.

Python Interface

Python interface is available at https://github.com/modsim/hopsy.

CMake options

Option Name	Default	Description
HOPS_HDF5_SUPPORT	OFF	Enables HDF5 support with HighFive. Use -DHOPS_BENCHMARKS=ON to enable.
HOPS_BENCHMARKS	OFF	Enables compilation of Benchmarks (Requires Celero). Use -DHOPS_BENCHMARKS=ON to enable.
HOPS_DOCS	ON	Enables generation of documentation. Use -DHOPS_DOCS=OFF to disable.
HOPS_BINARIES	ON	Enables compilation of hops executables. Use -DHOPS_EXAMPLES=OFF to disable.
HOPS_TESTS	ON	Enables compilation of unit tests. Use -DHOPS_TESTS=OFF to disable.
HOPS_LIBRARY_TYPE	SHARED	Type of library to build. Options are STATIC, SHARED or HEADER_ONLY (deprecated)

Install on Linux:

$ mkdir cmake-build-release cd cmake-build-release      # create and switch into 
                                                        # directory for out-of-source build
$ cmake .. -DCMAKE_BUILD_TYPE=Release                   # run cmake
$ make                                                  # build hops
$ make test                                             # run Tests
$ sudo make install                                     # install

Install on Windows 10:

Use an IDE (e.g. CLion) to parse the project and its CMakeLists.txt.

Examples

See the examples directory for demonstrations on how to use the library.

Supported Compilers

g++
Clang
Microsoft Visual C++

Troubleshooting

If you run into trouble finding CLP on Linux (e.g. Ubuntu 20.04), try extending the cmake prefix path:

-DCMAKE_PREFIX_PATH=/usr/lib/x86_64-linux-gnu
In case something went wrong fetching the git lfs content, try: git lfs pull