Data Science and Machine Learning (ML) methods are increasingly being used to transform the way research is being conducted in materials science to enable new discoveries and design new materials. For any materials science researcher or student, it may be daunting to figure out if ML techniques are useful for them or, if so, which ones are applicable in their individual contexts, and how to study the effectiveness of these methods systematically.
Machine Learning (ML) has evolved as a subfield of Artificial Intelligence (AI), learning from the data collected historically or from experiments, and using it for future actions. In general, ML models consider the patterns of the input and adjusts internal structures to approximate the relationship between input and output. ML is also used to identify hidden patterns of data distributions to come up with meaningful relationships. The ability to learn unforeseen relationships from data without depending on explicitly programmed prior guidance is one of the main reasons why there are a plethora of ML-based applications. The very early definition for ML, “Field of study that gives computers the ability to learn without being explicitly programmed” is still valid.
Key features:
Provides broad coverage of data science and ML fundamentals to materials science researchers so that they can confidently leverage these techniques in their research projects
Offers introductory material in topics such as ML, data integration, and 2D materials
Provides in-depth coverage of current ML methods for validating 2D materials using both experimental and simulation data, researching and discovering new 2D materials, and enhancing ML methods with physical properties of materials
Discusses customized ML methods for 2D materials data and applications and high-throughput data acquisition
Describes several case studies illustrating how ML approaches are currently leading innovations in the discovery, development, manufacturing, and deployment of 2D materials needed for strengthening industrial products
Gives future trends in ML for 2D materials, explainable AI, and dealing with extremely large and small, diverse datasets
Aimed at materials science researchers, this book allows readers to quickly, yet thoroughly, learn the ML and AI concepts needed to ascertain the applicability of ML methods in their research.
Author(s): Parvathi Chundi, Venkataramana Gadhamshetty, Bharat K. Jasthi, Carol Lushbough
Publisher: CRC Press
Year: 2023
Language: English
Pages: 249
Cover
Half Title
Title Page
Copyright Page
Table of Contents
Chapter 1 Introduction to Machine Learning for Analyzing Material–Microbe Interactions
1.1 Introduction
References
Chapter 2 Introduction to 2D Materials
2.1 Classification of 2D Materials
2.2 Synthesis of 2D Materials
2.2.1 Top-Down Methods
2.2.2 Bottom-Up Methods
2.2.3 Layer Transfer Methods
2.3 Functionality of 2D Materials
2.3.1 Mechanical Properties
2.3.2 Electrical Properties
2.3.3 Optical Properties
2.4 Applications of 2D Materials
References
Chapter 3 An Overview of Machine Learning
3.1 Introduction
3.1.1 The Processing Pipeline of an ML Task
3.1.2 Data Integration
3.1.3 Data Preparation
3.1.4 Model Building
3.1.5 Model Evaluation
3.2 ML Algorithms
3.2.1 Bias and Variance
3.3 Unsupervised Learning
3.3.1 Cluster Analysis
3.3.2 Principal Component Analysis (PCA)
3.4 Supervised Learning
3.4.1 Regression
3.4.2 Classification
3.4.3 Supervised Learning Variants: Self-Supervised Learning
3.5 Deep Learning
3.5.1 Convolutional Neural Networks (CNNs)
3.6 Recurrent Neural Networks (RNN)
References
Chapter 4 Discovery of 2D Materials with Machine Learning
4.1 Introduction: High-Throughput Screening
4.2 ML Approaches for 2D Materials Research
4.2.1 Three ML Approaches for 2D Materials Research
4.2.2 A Summary of the Use of Machine Learning in 2D Materials Research
4.3 Prediction of 2D Material Properties Using Machine Learning
4.4 Application Machine Learning Approaches to Discover Novel 2D Materials
4.4.1 Predictions of Crystal Structure
4.4.2 Prediction of Components
4.5 Machine Learning for Miscellaneous Functions
4.6 Assessment of Common Challenges and Their Prevention Methods
4.6.1 The Problems with Model Building
4.6.2 Usability
4.6.3 Learning Efficiency
4.7 Conclusions
References
Chapter 5 Bacterial Image Segmentation through Deep Learning Approach
5.1 Introduction
5.2 Literature Review and Related Work
5.2.1 Conventional Approaches for Semantic Segmentation
5.2.2 Contour-Based Methods
5.2.3 Ellipse-Fitting Methods
5.2.4 CNN-Based Approaches
5.3 Methodology
5.3.1 Data Collection
5.3.2 Image Preprocessing
5.3.3 ViTransUNet
5.4 Experimental Design and Results
5.4.1 Experimental Setup
5.4.2 Evaluation Metrics
5.4.3 Evaluation Results
5.5 Conclusion and Future Work
References
Chapter 6 Self-Supervised Learning-Based Classification of Scanning Electron Microscope Images of Biofilms
6.1 Introduction
6.2 Self-Supervised Learning for Image Analyses
6.2.1 Pretext Tasks
6.2.2 Downstream Tasks on Medical Imaging
6.3 Use of Super-Resolution to Address the Heterogeneity and Quality of SEM Biofilm Images
6.3.1 Methodology
6.3.2 Experimental Setup and Results
6.3.3 Summary
6.4 Classification of SEM Biofilms Using SSL
6.4.1 Dataset
6.4.2 Image Pre-Processing
6.4.3 Annotation, Patch Generation, and Object Masking
6.4.4 Self-Supervised Training
6.4.5 Downstream Task
6.4.6 Experiments
6.4.7 Evaluation
6.4.8 Results
6.4.9 Discussion
6.4.10 Summary
6.5 Conclusion
Acknowledgements
References
Chapter 7 Quorum Sensing Mechanisms, Biofilm Growth, and Microbial Corrosion Effects of Bacterial Species
Definitions
Acronyms
7.1 Introduction
7.2 Quorum Sensing
7.3 Key Quorum Sensing Molecules and Their Signaling Mechanisms
7.4 Quorum Sensing in Relation to Stress Response
7.5 Background on Biofilms with Focus on Its Ecology in Natural Ecosystems
7.6 Quorum Sensing, Biofilm Growth, and Microbiologically Influenced Corrosion
7.6.1 QS, Biofilm Growth, and MIC
7.6.2 Bioinformatics Analysis
7.7 Adhesion-Induced Emergent Properties in Biofilm
7.8 Methods to Inhibit Quorum Sensing
7.9 Conclusion
Acknowledgments
References
Chapter 8 Data-Driven 2D Material Discovery Using Biofilm Data and Information Discovery System (Biofilm-DIDS)
8.1 Introduction
8.1.1 Microbial Community, Biofilm, and Material–Biofilm Interaction
8.1.2 Complex System Design: SDLC and Agile Methodology Meets Big Data
8.1.3 Big Data Mining and Knowledge Discovery
8.2 Interface between the Living and the Non-Living: a System Thinking Approach
8.2.1 System Understanding of Biointerface
8.2.2 Big Data in Biointerfaces
8.3 Biofilm-DIDS Overview
8.4 Using Biofilm-DIDS to Extract Biocorrosion Gene of Interest from the Literature and Material Dimension Prediction
8.4.1 Expert Informed Relevant Dataset Extraction from User Free Text Question
8.4.2 Downstream Analysis for Material Dimension Prediction
8.5 Conclusions
Acknowledgments
References
Chapter 9 Machine Learning-Guided Optical and Raman Spectroscopy Characterization of 2D Materials
9.1 Introduction
9.2 Established Surface Characterization Techniques
9.3 ML-Guided Optical Detection of 2D Materials
9.4 ML-Guided Raman Spectroscopy Detection of 2D Materials
9.5 Common Challenges to ML in Raman Spectroscopy
9.6 Future Prospects
9.7 Summary
References
Chapter 10 Atomistic Experiments for Discovery of 2D Coatings: Biological Applications
10.1 Introduction
10.2 Molecular Dynamics (Algorithms and Methods)
10.2.1 Empirical Forcefields
10.2.2 Periodic Boundary Conditions
10.2.3 Binding Energy
10.2.4 Free Energy
10.2.5 Umbrella Sampling
10.2.6 Coarse-Grained Modeling
10.3 Employment of MD on Functional 2D Materials
10.3.1 Graphene and Its Structural Defects
10.3.2 The Emergence of Bioinformatics: Applications and Methodologies
10.3.3 Current Trends in Biomolecular Simulation and Modeling
10.4 Machine Learning
10.4.1 ML Methods for 2D Materials
10.4.2 ML for Force Field Development and Parameterization
10.4.3 ML for Protein Structure Prediction
10.5 Summary
References
Chapter 11 Machine Learning for Materials Science: Emerging Research Areas
11.1 Introduction
11.2 Applications of ML in Materials Science
11.2.1 Additive Manufacturing
11.2.2 Combinatorial Synthesis and Machine Learning-Assisted Discovery of Thin Films
11.2.3 Machine Learning-Assisted Properties Prediction of Bulk Alloys
11.2.4 Design of Drug-Releasing Materials with Machine Learning
11.2.5 AI and ML Tools for Search and Discovery of Quantum Materials
11.3 Gaps and Barriers to Implementation
References
Chapter 12 The Future of Data Science in Materials Science
12.1 Introduction
12.2 Learning with Small Training Datasets
12.2.1 Data Augmentation
12.2.2 Semisupervised Learning
12.2.3 Transfer Learning
12.2.4 Few-Shot Learning
12.3 Physics-Inspired Neural Networks
12.4 Digital Twins
12.5 Data-Centric Artificial Intelligence
12.5.1 Data Collection
12.5.2 Robust and Fair Model Training
12.5.3 Continuous Learning
12.6 GPT Models
12.7 Future Directions in Using ML for 2D Materials
References
Index
