Data-based methods for materials design and discovery : basic ideas and general methods / Ghanshyam Pilania (Los Alamos National Laboratory, Los Alamos, New Mexico), Prasanna V. Balachandran (University of Virginia, Charlottesville, Virgina), James E. Gubernatis (Santa Fe, New Mexico),Turab Lookman (Santa Fe, New Mexico)

By:

Pilania, Ghanshyam [VerfasserIn]

Contributor(s):

Resource type: Ressourcentyp: Buch (Online)Book (Online)Language: English Series: Synthesis lectures on materials and optics ; 1Publisher: [San Rafael, California] : Morgan & Claypool Publishers, [2020]Description: 1 Online-Ressource (xv, 172 page) : illustrations (some color)ISBN:

9781681737386

Subject(s):

Additional physical formats: 9781681737379. | 9781681737393. | Erscheint auch als: Data-based methods for materials design and discovery. Druck-Ausgabe [San Rafael, Calif.] : Morgan & Claypool, 2020. XV, 172 pagesDDC classification:

620.110285

LOC classification:

TA404.23

Online resources:

Zugang im Netz des KIT

Summary: Intro -- Preface -- Acknowledgments -- Introduction -- Scope and Plan -- Historical Perspective -- Machine Learning -- An Overview -- Bias vs. Variance Problem -- Toolboxes -- No Free Lunch Theorems -- References -- Materials Representations -- Conditions for a Valid Representation -- Hierarchy of Materials Representations -- Microscopic Representations -- Mesoscopic Representations -- Macroscopic Representations -- References -- Learning with Large Databases -- Databases -- Experimental Databases -- Computational Databases -- Combined Databases -- Materials Design and DiscoverySummary: Machine learning methods are changing the way we design and discover new materials. This book provides an overview of approaches successfully used in addressing materials problems (alloys, ferroelectrics, dielectrics) with a focus on probabilistic methods, such as Gaussian processes, to accurately estimate density functions. The authors, who have extensive experience in this interdisciplinary field, discuss generalizations where more than one competing material property is involved or data with differing degrees of precision/costs or fidelity/expense needs to be consideredSummary: Multivariate Gaussian Distributions -- Gaussian Processes -- Regression -- Classification -- References -- Authors' Biographies -- Blank PageSummary: Pareto Optimality -- Pareto Front -- Designing Materials Using Multi-Objective Optimization Strategies -- Multi-Objective Optimization Based on Forward and Inverse Modeling -- Multi-Objective Optimization Based on Optimal Learning -- References -- Multi-Fidelity Learning -- Kriging -- Co-Kriging -- Recursive Co-Kriging -- Multi-Fidelity Learning for Materials Design and Discovery -- References -- Some Closing Thoughts -- The Bayesian Perspective -- Tensors not Tables of Data -- Machine Learning in Experiment -- References -- Basic Notions of Probability TheorySummary: Substitution Probability Analysis -- T = 0 vs. T 0 K -- Ensemble Methods Based on Decision Trees -- Decision Trees -- Bagging and Random Forests -- Boosting and Gradient Tree Boosting -- Cross-Validation, Relative Feature Importance, and Partial Dependencies -- References -- Learning with Small Databases -- Background -- Bayes Theorem -- Gaussian Processes -- Applications -- Bayesian Global Optimization -- Examples of Utility Functions -- Applications to Materials Science -- References -- Multi-Objective Learning -- Introduction -- Definitions and Key Concepts -- Pareto DominancePPN: PPN: 1755158718Package identifier: Produktsigel: BSZ-4-NLEBK-KAUB | ZDB-4-NLEBK

Holdings ( 0 )

No physical items for this record