eth-cscs / PASC_inference

PASC_inference

HPC C++ library for signal denoising and clustering time-series. Based on FEM-H1 regularisation proposed by Illia Horenko et al. and supported by "PASC project: Towards the HPC-inference of causality networks from multiscale economical data".

Generated Doxyden documentation can be found here: http://eth-cscs.github.io/PASC_inference/

Installation

• download library

git clone https://github.com/eth-cscs/PASC_inference.git
cd PASC_inference

• download submodules

./util/update_submodules

• be sure that right modules are loaded and PETSc library is installed (and PETSC_DIR, PETSC_ARCH are properly set); if you are compiling the code on PIZ Daint, you can use scripts, which automatically set right modules and PETSc

source util/module_load_daint

• compile examples

cd examples
mkdir build
cd build
cmake -DFIND_PETSC=ON ..

(in case of CUDA implementation, use -DUSE_CUDA=ON)

• see the list of avaiable examples and choose which one to compile using -DTEST_...=ON

cmake -DTEST_SIGNAL1D=ON ..

• compile the library with examples

make