Generative AI for Science

Generative AI for Science

J. Paul Liu
This is a sample of the book's content.Buy on Leanpub

Table of Contents

Generative AI for Science

  • From the Author
    • A Journey from Classroom to Lab β€” and Now to You
    • Origins
    • What Makes This Book Different
    • How to Use This Book
    • What You Will Learn
    • A Note on Collaboration
    • Downloading all codes or slides
    • Prerequisites
    • AI Use Disclaimer
    • Writing Workflow
  • Chapter 1 β€” Generative AI: A New Frontier for Scientific Discovery
    • The New Frontier of Scientific Discovery
    • The AI Revolution in Scientific Discovery
    • What Makes Generative AI Different from Traditional ML?
    • Core Technologies Powering Generative AI
    • The Pre-Training Revolution
    • Generative AI Across Scientific Disciplines
    • Mathematical Foundations and Methods
    • Cross-Cutting Capabilities
    • A New Scientific Partner
    • The Path Forward
    • References and Further Readings
  • Chapter 2: Generative AI Fundamentals
    • Introduction: The Building Blocks of Generation
    • The Three Pillars of Generative AI
    • Part I: Transformers and Large Language Models
    • Part II: Diffusion Models and Flow Matching
    • Part III: VAEs and GANs
    • Part IV: Pre-Training and Fine-Tuning
    • Part V: Mathematical Foundations
    • Part VI: Types of Generative AI by Modality
    • Design Principles for Scientific Applications
    • Practical Considerations
    • Summary
    • References
  • Chapter 3: Scientific Data & Workflows
    • Introduction: The Data Challenge in Science
    • Part I: Unique Challenges of Scientific Data
    • Part II: Data Sources in Science
    • Part III: The FAIR Principles
    • Part IV: Data Preparation for AI
    • Part V: Integrating AI into Research Workflows
    • Part VI: Automated Workflow Generation
    • Summary
    • References
  • Chapter 4: Text, Code & Knowledge Generation for Scientists
    • Introduction: The Knowledge Synthesis Challenge
    • Part I: Literature Review and Synthesis
    • Part II: Retrieval-Augmented Generation (RAG)
    • Part III: Hypothesis Generation
    • Part IV: Code Generation for Research
    • Part V: Scientific Writing Assistance
    • Part VI: Educational Applications
    • Part VII: Domain-Specific LLM Systems
    • Part VIII: Limitations and Best Practices
    • Summary
    • References
  • Chapter 5: Data-to-Data Models
    • Introduction: The Data Scarcity Problem
    • Part I: Missing Data Imputation with Autoencoders
    • Part II: Synthetic Data Generation with GANs
    • Part III: Variational Autoencoders (VAEs)
    • Part IV: Gaussian Process Spatial Interpolation
    • Part V: Time Series Gap Filling
    • Summary and Key Takeaways
    • Next Steps
    • References
  • Chapter 6: Physics-Informed AI and Simulation
    • Introduction: Embedding Physics in Neural Networks
    • Part I: Physics-Informed Neural Networks (PINNs)
    • Part III Neural Network Surrogates for Simulations
    • Part IV: Code Optimization with AI
    • Part V: Automated Test Generation
    • Summary
    • References
  • Chapter 7: Domain Applications in Chemistry, Biology, Physics and Geoscience
    • Introduction: Generative AI Across the Sciences
    • Part I: Chemistry & Materials Science
    • Summary
    • References
    • Part II: Biology & Biomedicine
    • Summary: Biology & Biomedicine
    • References
    • Part III: Physics & Engineering
    • Summary: Physics & Engineering
    • References
    • Part IV: Geoscience & Climate Applications
    • Summary: Geoscience & Climate Applications
    • References
    • Part V: Cross-Cutting Applications in Deep Learning
    • Summary: Cross-Cutting Applications
    • References
  • Chapter 8: Fine-Tuning & Domain Adaptation
    • Introduction: Making General Models Domain-Specific
    • Part I: Why Fine-Tuning Works for Science
    • Part II: Parameter-Efficient Fine-Tuning (PEFT)
    • Part III: Practical Results - Biology Text Fine-Tuning
    • Part IV: Preparing Domain-Specific Training Data
    • Part V: Evaluation and Validation
    • Note: The dataset updates in real time, so each run may yield different results.
    • Part VI: Best Practices and Lessons Learned
    • Summary
    • References
  • Chapter 9: Multimodal Generative AI for Sciences
    • Introduction: Beyond Single-Modality AI
    • Part I: Vision-Language Models for Science
    • Part II: Graph-Text Models for Molecules
    • Part III: Time Series with Textual Context
    • Part IV: Multimodal Fusion Architectures
    • Part V: Scientific Document Understanding
    • Part VI: Training Multimodal Scientific Models
    • Part VII: Practical Applications
    • Summary
    • References
  • Chapter 10: Evaluation, Validation & Benchmarking
    • Introduction: Trust Through Rigorous Assessment
    • Part I: Core Evaluation Metrics
    • Part II: Validation Strategies
    • Part III: Benchmarking Datasets and Tasks
    • Part IV: Human Evaluation
    • Part V: Uncertainty Quantification
    • Part VI: Failure Analysis
    • Part VII: Robustness Testing
    • Summary
    • References
  • Chapter 11: Ethics & Responsible AI for Science
    • πŸ“– How to Use This Chapter
    • πŸ“Š Code Quick Reference
    • Introduction: The Unique Responsibility of Scientific AI
    • Part I: Reproducibility and Open Science
    • Part II: Bias and Fairness in Scientific AI
    • Part III: Environmental Impact of AI
    • Part IV: Dual-Use and Biosecurity
    • Part V: Data Privacy in Scientific Research
    • Part VI: Attribution and Scientific Integrity
    • Part VII: Equity and Access
    • Summary
    • References
  • Chapter 12: Deployment & MLOps for Scientific Applications
    • Introduction: From Notebooks to Production Science
    • Part I: Experiment Tracking & Management
    • Part II: Data Versioning & Lineage
    • Part III: Model Lifecycle Management
    • Part IV: Continuous Training Pipelines
    • Part V: Scientific Validation & Testing
    • Part VI: Deployment to Scientific Infrastructure
    • Part VII: Monitoring Production Models
    • References
  • Chapter 13: Future Directions & Conclusion
    • Introduction: Science at the Dawn of the AI Era
    • Part I β€” Emerging Architectures & Techniques
    • Part II β€” Multimodal Scientific AI
    • Part III β€” Foundation Models for Science
    • Part IV β€” AI for Scientific Reasoning
    • Part V β€” Open Challenges (Grouped)
    • Part VI β€” A Vision for the Next Decade
    • Part VII β€” Conclusion: The Scientific Method, Amplified
    • References
  • Acknowledgements
Generative AI for Science/Acknowledgements

Acknowledgements

https://leanpub.com/generativeaiforscience

In this chapter

  • Acknowledgements