| Session | ||
FS5 S1: Machine Learning Application to Spectroscopy and Imaging
| ||
| Presentations | ||
8:45am - 9:15am
Invited ID: 482 / FS5 S1: 1 Focused Sessions 5: Machine-Learning for Optics and Photonic Computing for AI Invited - Machine learning techniques for noise characterizatoin of optical frequeny combs Technical University of Denmark, Denmark We will present how widely used techniques such as autoencoders can be used to perform phase noise characterization of optical frequency combs. Optical frequency comb is an optical source that produces evenly spaced frequency lines. It is considered as a frequency ruler and has diverse applications in: frequency referencing, grid synchronization, optical communication, calibration for spectrographs e.g. 9:15am - 9:30am
ID: 279 / FS5 S1: 2 Focused Sessions 5: Machine-Learning for Optics and Photonic Computing for AI Distinguishing healthy and diseased chestnuts via THz spectroscopy and unsupervised learning 1Scuola Superiore Meridionale - Federico II, Italy; 2Istituto Nazionale di Ottica INO-CNR, Consiglio Nazionale delle Ricerche, Pozzuoli, Italy; 3ISASI, Institute of Applied Sciences and Intelligent Systems, Consiglio Nazionale delle Ricerche, Pozzuoli, Italy; 4IPSP, Istituto per la Protezione Sostenibile delle Piante, Consiglio Nazionale delle Ricerche, Portici, Italy Classifying chestnuts as healthy or diseased remains a complex challenge in quality assessment. In our study, we use THz imaging to determine accurately the health status of chestnuts. Through innovative spectroscopic analysis, we explore the potential of three distinct unsupervised data analysis techniques: Principal Component Analysis (PCA), k-Means Clustering, and Agglomerative Clustering. Compared to traditional analysis methods, our findings unveil the remarkable ability of these methods to differentiate between healthy, diseased and in an intermediate state chestnuts, even when concealed beneath the peel. This research not only advances our understanding of quality control in chestnut production but also highlights the potential of THz imaging in agricultural applications. 9:30am - 9:45am
ID: 372 / FS5 S1: 3 Focused Sessions 5: Machine-Learning for Optics and Photonic Computing for AI Physics-driven learning for digital holographic microscopy Université de Franche-Comté, CNRS, Institut FEMTO-ST, 25000 Besançon, France Deep neural networks based on physics-driven learning make it possible to train neural networks with a reduced data set and also have the potential to transfer part of the numerical computations to optical processing. The aim of this work is to develop the first deep holographic microscope device incorporating a hybrid neural network based on the plane-wave angular spectrum method for dynamic image autofocusing in microscopy applications. 9:45am - 10:00am
ID: 143 / FS5 S1: 4 Focused Sessions 5: Machine-Learning for Optics and Photonic Computing for AI Deep Classification from scattered light 1Center for Life Nano- & Neuroscience, Italian Institute of Technology; 2Institute of Nanotechnology of the National Research Council of Italy Photonic Stochastic Emergent Learning (PSEL) represents an innovative paradigm rooted in mathematical brain modelling and emergent memories. In this study, we explore the intersection of these concepts to address memory storage and classification tasks. Leveraging optical computing principles and random projections, PSEL constructs memory representations from the inherent randomness in nature. Specifically, we select a set of highly similar random states generated by coherent light scattered from a diffusive medium. Classification is performed by organizing the memories spatially into different classes and comparing inputs to those stored memories. The results demonstrate the efficacy of PSEL in memory construction and parallel classification, emphasizing its potential applications in high-performance computing and artificial intelligence systems. | ||