Where is the machine looking? Locating discriminative light-scattering features by class-activation mapping. (May 2020)

• Record Type: Journal Article
• Authors: Piedra, Patricio; Gobert, Christian; Kalume, Aimable; Pan, Yong-Le; Kocifaj, Miroslav; Muinonen, Karri; Penttilä, Antti; Zubko, Evgenij; Videen, Gorden
• Is Part Of: Journal of quantitative spectroscopy & radiative transfer. Volume 247(2020)
• Publication Date: 2020-05
• Subjects: Light scattering -- Deep learning; Spectrum analysis -- Periodicals; Radiation -- Periodicals; Analyse spectrale -- Périodiques; Rayonnement -- Périodiques; Radiation; Spectrum analysis; Periodicals; 543.0858
• DOI: 10.1016/j.jqsrt.2020.106936
• Languages: English
• ISSNs: 0022-4073

Abstract

Highlights: Used class-activation maps (CAM) to quantify the discriminative importance of light-scattering angles for our trained neural networks. The accuracy of our previous neural network was improved significantly using batch normalization, max-pooling, and leaky-ReLU activation function. Azimuthally-averaged CAMs highlighted scattering angles with high discriminative importance for specific particle classes and configuration of Mueller-matrix data input. CAMs can be used to inform scientists where to place detectors for strongest classification profile depending on the type of particles and measurement configuration. Abstract: We explore a technique called class-activation mapping (CAM) to investigate how a Machine Learning (ML) architecture learns to classify particles based on their light-scattering signals. We release our code, and also find that different regions of the light-scattering signals play different roles in ML classification. These regions depend on the type of particles being classified and on the nature of the data obtained and trained. For instance, the Mueller-matrix elements S 11 *, S 12 * and S 21 * had the greatest classification activation in the diffraction region. Linear polarization elements S 12 * and S 21 * were most accurate in the backscattering region for clusters of spheres and spores, and was most accurate in the diffraction region for other particle classes. The CAM technique was able to highlight light-scattering angles that maximize the potential for discrimination of similar particle classes. Such information is useful for designing detector systems to classify particles where limited space or resources are available, including flow cytometry and satellite remote sensing.