Speaker
Description
The quest to interpret spectra (UV, VIS, Raman, etc.) is a well-known problem for physicists from fields such as experimental, complex systems, or phonics, where with each sample a spectrum can be measured to check its quality and properties, that quest can be seen with each newly studied material or compound [1]. Recent advancements in theoretical frameworks, such as genetic algorithms, and series-based$^{1}$ Machine Learning [2], have provided tantalizing glimpses into the basic understanding of such spectroscopy and the nature of some nuances not yet well understood [3].
This presentation will introduce and explore a basic understanding of Machine Learning and inverse problems based on spectroscopy (absorption, scattering, and reflection). We will also delve into the implications of recent developments in Machine Learning-based inverse problems solved in selected spectroscopy techniques.
$^1$ It is often that when working with a series of data, that data is entangled with time, and thus it is commonly known as "time-series" data, and machine learning models developed upon it are known as time-series machine learning models. The spectrum might vary in a base for obtaining data and that is why the name "series-based" was used to minimize confusion, but techniques that might be used to analyze it with machine learning approaches are indeed known under the name "time-series".
References
[1] Tanudji, Jeffrey et al. (2023). Surface Facet Effect on the Adsorption of Iodine and Astatine on Gold Surface. e-Journal of Surface Science and Nanotechnology. 22.
[2] Mayer, Alexandre et al. “Genetic-algorithm-aided ultra-broadband perfect absorbers using plasmonic metamaterials.” Optics express 30 2 (2021): 1167-1181.
[3] Shizhao Lu, and Arthi Jayaraman. "Machine learning for analyses and automation of structural characterization of polymer materials".Progress in Polymer Science 153 (2024): 101828.
