☆ 4.7 Article

Integrating Chemical Information into Reinforcement Learning for Enhanced Molecular Geometry Optimization

JOURNAL OF CHEMICAL THEORY AND COMPUTATION (2023)

相关参考文献

注意：仅列出部分参考文献，下载原文获取全部文献信息。

Article Chemistry, Physical

A spur to molecular geometry optimization: Gradient-enhanced universal kriging with on-the-fly adaptive ab initio prior mean functions in curvilinear coordinates

Chong Teng et al.

Summary: We propose a molecular geometry optimization algorithm based on gradient-enhanced universal kriging (GEUK) with ab initio prior mean functions, integrating prior physical knowledge into surrogate-based optimization. Our implementation is both general and flexible, allowing optimizations in both Cartesian and curvilinear coordinates. We show that the GEUK optimizer accelerates geometry optimization in internal coordinates and demonstrate its efficiency in optimizing challenging molecules with high-accuracy electronic structure methods.

JOURNAL OF CHEMICAL PHYSICS (2023)

添加到收藏夹

Article Chemistry, Multidisciplinary

mad-GP: automatic differentiation of Gaussian processes for molecules and materials

Daniel Huang et al.

Summary: This paper introduces a Python library called mad-GP which simplifies the exploration of Gaussian process surrogate models for modeling potential energy surfaces (PESs). The library allows users to specify the prior mean function and kernel function, and handles all required derivatives using automatic differentiation. The design of mad-GP is validated through its application to geometry optimization of small molecules. The results demonstrate comparable performance between GPs that fit forces and those that fit both energies and forces, while also identifying other potential candidates for mean functions and kernels.

JOURNAL OF MATHEMATICAL CHEMISTRY (2022)

添加到收藏夹

Article Chemistry, Medicinal

Exploring Potential Energy Surfaces Using Reinforcement Machine Learning

Alexis W. Mills et al.

Summary: Reinforcement machine learning is implemented to survey potential energy surfaces and identify the global minima point. The use of an optimized reward function and physically motivated actions allows the agent to make advanced decisions and converge to an optimal solution more rapidly. This study demonstrates the feasibility of using reinforcement machine learning to solve complex optimization problems in chemistry.

JOURNAL OF CHEMICAL INFORMATION AND MODELING (2022)

添加到收藏夹

Article Chemistry, Physical

Dual-Level Training of Gaussian Processes with Physically Inspired Priors for Geometry Optimizations

Chong Teng et al.

Summary: This article investigates the effectiveness of Gaussian process regression in molecular geometry optimization and designs two types of physically inspired prior mean functions. Experimental results show that these prior mean functions can reduce the optimization steps by a factor of 2-3 compared to previous models. The study also highlights the importance of incorporating domain knowledge in Gaussian process to improve model performance.

JOURNAL OF CHEMICAL THEORY AND COMPUTATION (2022)

添加到收藏夹

Article Chemistry, Physical

Rational synthesis of ruthenium-based metallo-supramolecular polymers as heterogeneous catalysts for catalytic transfer hydrogenation of carbonyl compounds

Zi-Jie Gong et al.

Summary: In this study, ruthenium-based metallo-supramolecular polymers (Ru-MSPs) were synthesized and used for heterogeneously catalytic transfer hydrogenation of carbonyl compounds. The polymerization process was controlled by adjusting the ligand/metal ratio and synthesis conditions. The coordinatively unsaturated ruthenium atoms were identified as the active sites for carbonyl reduction, and the Ru-MSPs exhibited enhanced activity and reusability compared to monomeric counterparts.

APPLIED CATALYSIS B-ENVIRONMENTAL (2022)

添加到收藏夹

Article Chemistry, Physical

Comparative Analysis of Uncoupled Mode Approximations for Molecular Thermochemistry and Kinetics

Shih-Cheng Li et al.

Summary: Accurately predicting thermochemistry and kinetic parameters is crucial for reaction modeling. By improving the representation of anharmonic potential energy surfaces through uncoupled mode approximations, this study demonstrates the importance of properly defining one-dimensional potentials for increased accuracy. The results highlight the efficacy of sampling torsional modes separately from vibrational modes, and the advantages of using hybrid or translation-rotation-internal coordinate systems.

JOURNAL OF CHEMICAL THEORY AND COMPUTATION (2022)

添加到收藏夹

Article Multidisciplinary Sciences

Discovering faster matrix multiplication algorithms with reinforcement learning

Alhussein Fawzi et al.

Summary: Improving the efficiency of fundamental computations, such as matrix multiplication, is crucial for various fields, and machine learning can automate the discovery of efficient algorithms that surpass human-designed ones. AlphaTensor, based on deep reinforcement learning and AlphaZero, successfully discovered algorithms that outperform state-of-the-art ones. Its flexibility in accelerating algorithm discovery and optimization is showcased through different use-cases.

NATURE (2022)

添加到收藏夹

Article Chemistry, Physical

pysisyphus: Exploring potential energy surfaces in ground and excited states

Johannes Steinmetzer et al.

Summary: This study presents an external optimizer called pysisyphus, which can localize stationary points in both ground state and excited state of chemical reactions, providing multiple state-tracking algorithms. This tool offers all necessary tools for calculating reaction paths, starting from optimizing reactants to intrinsic reaction coordinate calculation.

INTERNATIONAL JOURNAL OF QUANTUM CHEMISTRY (2021)

添加到收藏夹

Article Nanoscience & Nanotechnology

Proton-Conductive Cerium-Based Metal-Organic Frameworks

Wei Huan Ho et al.

Summary: This study demonstrates that replacing Zr with Ce in MOF-808 significantly increases proton conductivity and reduces activation energy. Ce-MOF-808 stands out in proton conduction, suggesting that cerium-based MOFs could be a better platform for designing proton conductors in future energy-related applications.

ACS APPLIED MATERIALS & INTERFACES (2021)

添加到收藏夹

Article Chemistry, Physical

Learning to Optimize Molecular Geometries Using Reinforcement Learning

Kabir Ahuja et al.

Summary: This study introduces a reinforcement learning approach to improve the performance of quasi-Newton methods in molecular geometry optimization when initial guess geometries are far from local minima. The proposed model produces a correction term for the quasi-Newton step, calculated with the BFGS algorithm, reducing the optimization steps by about 30% for molecules with perturbed geometries. The new method shows similar convergence to BFGS with a line search procedure, but is faster due to avoiding multiple gradient evaluations.

JOURNAL OF CHEMICAL THEORY AND COMPUTATION (2021)

添加到收藏夹

Article Chemistry, Physical

Restricted-Variance Constrained, Reaction Path, and Transition State Molecular Optimizations Using Gradient-Enhanced Kriging

Ignacio Fdez Galvan et al.

Summary: Gaussian process regression has shown better performance than standard methods in molecular equilibrium geometry optimization, especially in constrained optimizations and transition states. The newly developed method outperforms the current standard in efficiency and robustness.

JOURNAL OF CHEMICAL THEORY AND COMPUTATION (2021)

添加到收藏夹

Review Chemistry, Multidisciplinary

Quantum Mechanical Calculations for Biomass Valorization over Metal-Organic Frameworks (MOFs)

Jyun-Yi Yeh et al.

Summary: Metal-organic framework (MOF) technology shows great potential in the valorization of biomass, and quantum mechanical simulations can provide insights into the catalytic processes, potentially facilitating the development of novel MOF-based biomass materials.

CHEMISTRY-AN ASIAN JOURNAL (2021)

添加到收藏夹

Article Chemistry, Physical

Geometry Optimization in Internal Coordinates Based on Gaussian Process Regression: Comparison of Two Approaches

Daniel Born et al.

Summary: This study extended the geometry optimization method based on Gaussian process regression to internal coordinates, showing that predicting gradients in Cartesian or internal coordinates can both reduce the number of optimization steps and provide efficiency and robustness on various molecular structures.

JOURNAL OF CHEMICAL THEORY AND COMPUTATION (2021)

添加到收藏夹

Review Chemistry, Physical

Understanding the Catalytic Activity of Microporous and Mesoporous Zeolites in Cracking by Experiments and Simulations

Shih-Cheng Li et al.

Summary: This review explores recent investigations on hydrocarbon cracking and secondary reactions in micro- and mesoporous zeolites, with a focus on the effects of different porous environments and active site structures on alkane adsorption and activation at the molecular level. The pros and cons of various computational methods used for zeolite simulations are also discussed.

CATALYSTS (2021)

添加到收藏夹

Article Chemistry, Physical

Geometry optimization using Gaussian process regression in internal coordinate systems

Ralf Meyer et al.

JOURNAL OF CHEMICAL PHYSICS (2020)

添加到收藏夹

Article Chemistry, Physical

Restricted-Variance Molecular Geometry Optimization Based on Gradient-Enhanced Kriging

Gerardo Raggi et al.

JOURNAL OF CHEMICAL THEORY AND COMPUTATION (2020)

添加到收藏夹

Article Chemistry, Physical

GFN2-xTB-An Accurate and Broadly Parametrized Self-Consistent Tight-Binding Quantum Chemical Method with Multipole Electrostatics and Density-Dependent Dispersion Contributions

Christoph Bannwarth et al.

JOURNAL OF CHEMICAL THEORY AND COMPUTATION (2019)

添加到收藏夹

Article Chemistry, Physical

Gaussian Process Regression for Minimum Energy Path Optimization and Transition State Search

Alexander Denzel et al.

JOURNAL OF PHYSICAL CHEMISTRY A (2019)

添加到收藏夹

Article Biochemistry & Molecular Biology

DrugBank 5.0: a major update to the DrugBank database for 2018

David S. Wishart et al.

NUCLEIC ACIDS RESEARCH (2018)

添加到收藏夹

Article Chemistry, Physical

Gaussian process regression for geometry optimization

Alexander Denzel et al.

JOURNAL OF CHEMICAL PHYSICS (2018)

添加到收藏夹

Article Chemistry, Physical

SchNet - A deep learning architecture for molecules and materials

K. T. Schuett et al.

JOURNAL OF CHEMICAL PHYSICS (2018)

添加到收藏夹

Proceedings Paper Computer Science, Artificial Intelligence

Improving L-BFGS Initialization For Trust-Region Methods In Deep Learning

Jacob Rafati et al.

2018 17TH IEEE INTERNATIONAL CONFERENCE ON MACHINE LEARNING AND APPLICATIONS (ICMLA) (2018)

添加到收藏夹

Article Chemistry, Physical

Geometry optimization made simple with translation and rotation coordinates

Lee-Ping Wang et al.

JOURNAL OF CHEMICAL PHYSICS (2016)

添加到收藏夹

Article Multidisciplinary Sciences

Human-level control through deep reinforcement learning

Volodymyr Mnih et al.

NATURE (2015)

添加到收藏夹

Article Multidisciplinary Sciences

Quantum chemistry structures and properties of 134 kilo molecules

Raghunathan Ramakrishnan et al.

SCIENTIFIC DATA (2014)

添加到收藏夹

Review Chemistry, Applied

Quantum Mechanical Modeling of Catalytic Processes

Alexis T. Bell et al.

ANNUAL REVIEW OF CHEMICAL AND BIOMOLECULAR ENGINEERING, VOL 2 (2011)

添加到收藏夹

Review Chemistry, Multidisciplinary

Geometry optimization

H. Bernhard Schlegel

WILEY INTERDISCIPLINARY REVIEWS-COMPUTATIONAL MOLECULAR SCIENCE (2011)

添加到收藏夹

Article Chemistry, Physical

Long-range corrected hybrid density functionals with damped atom-atom dispersion corrections

Jeng-Da Chai et al.

PHYSICAL CHEMISTRY CHEMICAL PHYSICS (2008)

添加到收藏夹

Article Chemistry, Physical

The M06 suite of density functionals for main group thermochemistry, thermochemical kinetics, noncovalent interactions, excited states, and transition elements: two new functionals and systematic testing of four M06-class functionals and 12 other functionals

Yan Zhao et al.

THEORETICAL CHEMISTRY ACCOUNTS (2008)

添加到收藏夹

Article Physics, Multidisciplinary

Structural relaxation made simple

Erik Bitzek et al.

PHYSICAL REVIEW LETTERS (2006)

添加到收藏夹

Article Computer Science, Software Engineering

Cubic regularization of Newton method and its global performance

Yurii Nesterov et al.

MATHEMATICAL PROGRAMMING (2006)

添加到收藏夹

Review Chemistry, Multidisciplinary

Exploring potential energy surfaces for chemical reactions: An overview of some practical methods

HB Schlegel

JOURNAL OF COMPUTATIONAL CHEMISTRY (2003)

添加到收藏夹

Article Chemistry, Physical

Methods for optimizing large molecules - Part III. An improved algorithm for geometry optimization using direct inversion in the iterative subspace (GDIIS)

O Farkas et al.

PHYSICAL CHEMISTRY CHEMICAL PHYSICS (2002)

添加到收藏夹

Article Chemistry, Physical

Linear scaling geometry optimisation and transition state search in hybrid delocalised internal coordinates

SR Billeter et al.

PHYSICAL CHEMISTRY CHEMICAL PHYSICS (2000)

添加到收藏夹

© Peeref 2019-2024. All rights reserved.