Cover Image

Machine learning in chemistry : the impact of artificial intelligence /

Show other versions (1)

This book provides practical examples of machine learning applied to science to help researchers make an informed choice about using the method in chemistry.

Bibliographic Details
Other Authors: Cartwright, Hugh M., 1948- (Editor)
Format: eBook
Language:English
Published: Cambridge : Royal Society of Chemistry, 2020.
Series:Theoretical and computational chemistry ; 17.
Subjects:
Chemistry > Data processing.
Machine learning.
Electronic books.
Online Access:Connect to the full text of this electronic book
Table of Contents:
  • Intro
  • Title
  • Copyright
  • Preface
  • Contents
  • Chapter 1 Computers as Scientists
  • 1.1 What Is Computational Science?
  • 1.2 What Is Artificial Intelligence?
  • 1.3 What Is Machine Learning?
  • References
  • Chapter 2 How Do Machines Learn?
  • 2.1 Framing the Question
  • 2.2 Gathering Data
  • 2.3 Setting up the Algorithm
  • 2.4 The Training Procedure
  • 2.5 Overcoming Flaws
  • 2.6 Deploying the Algorithms
  • Note on References
  • References
  • Chapter 3 MedChemInformatics: An Introduction to Machine Learning for Drug Discovery
  • 3.1 Introduction
  • 3.1.1 Artificial Intelligence vs. Machine Learning
  • 3.1.2 Supervised vs. Unsupervised Learning
  • 3.2 Forewarned Is Forearmed
  • 3.2.1 Dataset Assembly and Curation
  • 3.2.2 Model Building
  • 3.2.3 Principal Component Analysis
  • 3.2.4 When Machines Learn in 3D, Alignment IS EVERYTHING
  • 3.2.5 Activity Range and Distribution of Activities for QSAR Modelling
  • 3.2.6 Outliers
  • 3.3 Deep Learning and Neural Networks
  • 3.4 Underlying Mathematics
  • 3.4.1 "Observables" and "Features"
  • 3.4.1.1 Molecular Descriptors and Fingerprints
  • 3.4.2 Comparing Molecules Based on Their Descriptors
  • 3.4.2.1 Tanimoto Similarity
  • 3.4.2.2 Tversky Similarity
  • 3.4.2.3 Dice Similarity
  • 3.4.2.4 Which Similarity Metric Works Best?
  • 3.4.3 Model Quality and Statistics
  • 3.4.3.1 Coefficients of Determination
  • 3.4.3.2 Kendall's Tau
  • 3.4.4 Overtraining and Characteristics of Good Descriptors
  • 3.5 Machine Learning Methods
  • 3.5.1 k-Nearest Neighbours
  • 3.5.2 Linear Regression
  • 3.5.2.1 Ordinary or Partial Least Squares?
  • 3.5.3 Decision Trees and Random Forests
  • 3.5.3.1 Information
  • 3.5.3.2 Algorithms
  • 3.5.3.3 Decision Tree Ensembles
  • 3.5.4 Support Vector Machines
  • 3.5.4.1 Hyperplanes
  • 3.5.4.2 Kernels
  • 3.5.4.3 Support Vectors
  • 3.6 Summary
  • Acknowledgements
  • References
  • Chapter 4 Machine Learning for Nonadiabatic Molecular Dynamics
  • 4.1 Introduction
  • 4.2 Methods
  • 4.2.1 Machine Learning (ML) Models
  • 4.2.1.1 Linear Model (LR)
  • 4.2.1.2 Kernel Ridge Regression (KRR)
  • 4.2.1.3 Support Vector Regression (SVR)
  • 4.2.1.4 Neural Networks (NNs)
  • 4.2.1.5 Training Process
  • 4.2.1.6 Descriptors
  • 4.2.1.7 Distance-matrix Based Descriptors
  • 4.2.1.8 FCHL: Faber-Christensen-Huang-Lilienfeld Representation
  • 4.2.2 Training Set Generation
  • 4.2.2.1 Initial Training Set
  • 4.2.2.2 Adaptive Sampling for Excited States
  • 4.2.3 Wave Function Phases
  • 4.2.4 Phase Correction Algorithm
  • 4.2.5 Surface Hopping Dynamics
  • 4.3 Example: Methylenimmonium Cation
  • 4.3.1 ML Surface Hopping Dynamics
  • 4.3.2 Energy Conservation
  • 4.3.3 Further Tools of ML Models
  • 4.4 Conclusion and Outlook
  • References
  • Chapter 5 Machine Learning in Science
  • A Role for Mechanical Sympathy?
  • 5.1 Introduction
  • 5.1.1 A Little History
  • 5.1.2 Challenges
  • 5.2 Problems and Solutions
  • 5.2.1 How Many Samples do I Need to Train an AI?