Automatic Feature Selection for Atom-Centered Neural Network Potentials Using a Gradient Boosting Decision Algorithm

被引:0
|
作者
Li, Renzhe [1 ]
Wang, Jiaqi [1 ]
Singh, Akksay [1 ,2 ,3 ]
Li, Bai [1 ]
Song, Zichen [1 ,4 ]
Zhou, Chuan [1 ]
Li, Lei [1 ]
机构
[1] Southern Univ Sci & Technol, Dept Mat Sci & Engn, Shenzhen 518055, Peoples R China
[2] Univ Texas Austin, Dept Chem, Austin, TX 78712 USA
[3] Univ Texas Austin, Inst Computat Engn & Sci, Austin, TX 78712 USA
[4] City Univ Hong Kong, Dept Mat Sci & Engn, Kowloon, Hong Kong 999077, Peoples R China
基金
国家重点研发计划; 中国国家自然科学基金;
关键词
LIQUID-METAL; FORCE-FIELD; APPROXIMATION; DYNAMICS; PERFORMANCE; SIMULATION; MODEL;
D O I
10.1021/acs.jctc.4c01176
中图分类号
O64 [物理化学(理论化学)、化学物理学];
学科分类号
070304 ; 081704 ;
摘要
Atom-centered neural network (ANN) potentials have shown high accuracy and computational efficiency in modeling atomic systems. A crucial step in developing reliable ANN potentials is the proper selection of atom-centered symmetry functions (ACSFs), also known as atomic features, to describe atomic environments. Inappropriate selection of ACSFs can lead to poor-quality ANN potentials. Here, we propose a gradient boosting decision tree (GBDT)-based framework for the automatic selection of optimal ACSFs. This framework takes uniformly distributed sets of ACSFs as input and evaluates their relative importance. The ACSFs with high average importance scores are selected and used to train an ANN potential. We applied this method to the Ge system, resulting in an ANN potential with root-mean-square errors (RMSE) of 10.2 meV/atom for energy and 84.8 meV/& Aring; for force predictions, utilizing only 18 ACSFs to achieve a balance between accuracy and computational efficiency. The framework is validated using the grid searching method, demonstrating that ACSFs selected with our framework are in the optimal region. Furthermore, we also compared our method with commonly used feature selection algorithms. The results show that our algorithm outperforms the others in terms of effectiveness and accuracy. This study highlights the significance of the ACSF parameter effect on the ANN performance and presents a promising method for automatic ACSF selection, facilitating the development of machine learning potentials.
引用
收藏
页码:10564 / 10573
页数:10
相关论文
共 41 条
  • [1] Pair-distribution-function guided optimization of fingerprints for atom-centered neural network potentials
    Li, Lei
    Li, Hao
    Seymour, Ieuan D.
    Koziol, Lucas
    Henkelman, Graeme
    JOURNAL OF CHEMICAL PHYSICS, 2020, 152 (22): : 224102
  • [2] High-dimensional neural network potentials for magnetic systems using spin-dependent atom-centered symmetry functions
    Eckhoff, Marco
    Behler, Joerg
    NPJ COMPUTATIONAL MATERIALS, 2021, 7 (01)
  • [3] Solar Flare Prediction and Feature Selection Using a Light-Gradient-Boosting Machine Algorithm
    Vysakh, P. A.
    Mayank, Prateek
    SOLAR PHYSICS, 2023, 298 (11)
  • [4] Attentional Neural Network: Feature Selection Using Cognitive Feedback
    Wang, Qian
    Zhang, Jiaxing
    Song, Sen
    Zhang, Zheng
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 27 (NIPS 2014), 2014, 27
  • [5] A new feature selection algorithm and composite neural network for electricity price forecasting
    Keynia, Farshid
    ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE, 2012, 25 (08) : 1687 - 1697
  • [6] Predicting Ice Phenomena in a River Using the Artificial Neural Network and Extreme Gradient Boosting
    Graf, Renata
    Kolerski, Tomasz
    Zhu, Senlin
    RESOURCES-BASEL, 2022, 11 (02):
  • [7] Construction of high-dimensional neural network potentials using environment-dependent atom pairs
    Jose, K. V. Jovan
    Artrith, Nongnuch
    Behler, Joerg
    JOURNAL OF CHEMICAL PHYSICS, 2012, 136 (19):
  • [8] Wind speed forecasting using deep neural network with feature selection
    Liu, Xiangjie
    Zhang, Hao
    Kong, Xiaobing
    Lee, Kwang Y.
    NEUROCOMPUTING, 2020, 397 : 393 - 403
  • [9] Predicting the surfactant-polymer flooding performance in chemical enhanced oil recovery: Cascade neural network and gradient boosting decision tree
    Larestani, Aydin
    Mousavi, Seyed Pezhman
    Hadavimoghaddam, Fahimeh
    Ostadhassan, Mehdi
    Hemmati-Sarapardeh, Abdolhossein
    ALEXANDRIA ENGINEERING JOURNAL, 2022, 61 (10) : 7715 - 7731
  • [10] Time series prediction using PSO-optimized neural network and hybrid feature selection algorithm for IEEE load data
    Sheikhan, Mansour
    Mohammadi, Najmeh
    NEURAL COMPUTING & APPLICATIONS, 2013, 23 (3-4): : 1185 - 1194