Scholarlink

Recent years have witnessed a tremendous expansion of Artificial Intelligence (AI) in both technology and science. Driven by the increasing computational power of computers, AI has dominated fields in image processing, internet technologies, automation, drug discovery, and many others. Moreover, AI has permeated popular culture and become a subject of philosophical discussion, particularly after the development of large language models (LLMs) that can convincingly mimic human communication. However, it is essential to recognize the complexity and diversity of technologies encompassed by the term “AI,” which was coined in the 1950s. While AI is often associated with artificial neural networks and large language models, it encompasses a plethora of different concepts varying in complexity. The applicability of these techniques depends on the nature of the problem and requires not only programming skills but also knowledge of data science and the phenomena being investigated. This is particularly true in LIBS, where problems can significantly differ in complexity and data availability. The aim of this chapter is to introduce a range of AI-associated techniques that are valuable for LIBS, depending on the application. These include not only shallow and deep machine learning models, which are currently primarily identified with AI, but also data preprocessing and preparation tools that are crucial in numerous scientific and technical problems addressed by LIBS. After defining the necessary terms and techniques, they will be described in more detail to prepare readers for their application in their research. This approach helps to avoid confusion arising from the multitude of names encountered by machine learning enthusiasts during their initial exposure to this discipline. References to scientific literature on ML in LIBS will assist in identifying problem classes that are most relevant for specific ML models. The chapter concludes with reflections on the future of AI in LIBS, supported by a practical presentation of relevant software tools in the appendix.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Data Processing, Shallow and Deep Learning Models for LIBS: Basics, Applications, and Tools

  • 章节

  • 发表日期：

    2025年

查看全文
  • Paweł Gąsior

摘要

Recent years have witnessed a tremendous expansion of Artificial Intelligence (AI) in both technology and science. Driven by the increasing computational power of computers, AI has dominated fields in image processing, internet technologies, automation, drug discovery, and many others. Moreover, AI has permeated popular culture and become a subject of philosophical discussion, particularly after the development of large language models (LLMs) that can convincingly mimic human communication. However, it is essential to recognize the complexity and diversity of technologies encompassed by the term “AI,” which was coined in the 1950s. While AI is often associated with artificial neural networks and large language models, it encompasses a plethora of different concepts varying in complexity. The applicability of these techniques depends on the nature of the problem and requires not only programming skills but also knowledge of data science and the phenomena being investigated. This is particularly true in LIBS, where problems can significantly differ in complexity and data availability. The aim of this chapter is to introduce a range of AI-associated techniques that are valuable for LIBS, depending on the application. These include not only shallow and deep machine learning models, which are currently primarily identified with AI, but also data preprocessing and preparation tools that are crucial in numerous scientific and technical problems addressed by LIBS. After defining the necessary terms and techniques, they will be described in more detail to prepare readers for their application in their research. This approach helps to avoid confusion arising from the multitude of names encountered by machine learning enthusiasts during their initial exposure to this discipline. References to scientific literature on ML in LIBS will assist in identifying problem classes that are most relevant for specific ML models. The chapter concludes with reflections on the future of AI in LIBS, supported by a practical presentation of relevant software tools in the appendix.