Assessing Gaussian Process Regression and Permutationally Invariant Polynomial Approaches To Represent High-Dimensional Potential Energy Surfaces

2018 ◽  
Vol 14 (7) ◽  
pp. 3381-3396 ◽  
Author(s):  
Chen Qu ◽  
Qi Yu ◽  
Brian L. Van Hoozen ◽  
Joel M. Bowman ◽  
Rodrigo A. Vargas-Hernández
Keyword(s):  
Potential Energy ◽  
Gaussian Process ◽  
Potential Energy Surfaces ◽  
Gaussian Process Regression ◽  
High Dimensional ◽  
Invariant Polynomial ◽  
Energy Surfaces

scholarly journals Easy representation of multivariate functions with low-dimensional terms via Gaussian process regression kernel design: applications to machine learning of potential energy surfaces and kinetic energy densities from sparse data

2022 ◽  
Author(s):  
Sergei Manzhos ◽  
Eita Sasaki ◽  
Manabu Ihara
Keyword(s):  
Machine Learning ◽  
Kinetic Energy ◽  
Potential Energy ◽  
Gaussian Process ◽  
Potential Energy Surfaces ◽  
Gaussian Process Regression ◽  
Multivariate Functions ◽  
Low Dimensional ◽  
Energy Surfaces ◽  
Kernel Design

Abstract We show that Gaussian process regression (GPR) allows representing multivariate functions with low-dimensional terms via kernel design. When using a kernel built with HDMR (High-dimensional model representation), one obtains a similar type of representation as the previously proposed HDMR-GPR scheme while being faster and simpler to use. We tested the approach on cases where highly accurate machine learning is required from sparse data by fitting potential energy surfaces and kinetic energy densities.

Calculations of the active mode and energetic barrier to electron attachment to CF3 and comparison with kinetic modeling of experimental results

2016 ◽  
Vol 18 (45) ◽  
pp. 31064-31071 ◽  
Author(s):  
Huixian Han ◽  
Benjamin Alday ◽  
Nicholas S. Shuman ◽  
Justin P. Wiens ◽  
Jürgen Troe ◽  
...  
Keyword(s):  
Neural Network ◽  
Potential Energy ◽  
Ab Initio ◽  
Kinetic Modeling ◽  
Potential Energy Surfaces ◽  
Electron Attachment ◽  
Invariant Polynomial ◽  
Active Mode ◽  
Polynomial Neural Network ◽  
Energy Surfaces

Six-dimensional potential energy surfaces of both CF3 and CF3− were developed by fitting ∼3000 ab initio points using the permutation invariant polynomial-neural network (PIP-NN) approach.

Sign in / Sign up

Export Citation Format

Share Document