/Chem-Graph-Kernel-Machine

Machine Learning using marginalized graph kernel for chemical molecules.

Primary LanguagePythonMIT LicenseMIT

Chem-Graph-Kernel-Machine

Predicting molecular properties using Marginalized Graph Kernel, GraphDot.

It supports regression (GPR) and classification (GPC, SVM) tasks on

  • pure compounds.
  • mixtures.

Besides molecular graph, additional vector could also be added as input, such as temperature, pressure, etc.

Predicting Single-Substance Phase Diagrams: A Kernel Approach on Graph Representations of Molecules.

A Comparative Study of Marginalized Graph Kernel and Message-Passing Neural Network.

Installation

GCC (7.*), NVIDIA Driver and CUDA toolkit(>=10.1). Python 3.10 is suggested.

pip install numpy==1.22.3 git+https://gitlab.com/Xiangyan93/graphdot.git@feature/xy git+https://github.com/bp-kelley/descriptastorus typed-argument-parser mgktools==1.1.0

For some combinations of GCC and CUDA, only old version of pycuda workspip install pycuda==2020.1

Usages

  1. The executable files are in directory run.
  2. The hyperparameter files in json format are placed in directory hyperparameters.