Practical applications of neural network with Kolmogorov-Arnold layers in spectral analysis

Abstract The application of a neural network based on the CNN-KAN architecture for the automated analysis of spectroscopic data is presented. The study focuses on issues related to food quality assessment and the detection of different types of plastics used in packaging. The results demonstrate that the CNN-KAN architecture that combines convolutional and Kolmogorov-Arnold layers can be successfully applied to both binary and multiclass classification tasks, including scenarios involving datasets with a limited number of samples and/or low-resolution spectra. The study encompasses a range of practical applications, including the detection of pur\'ee samples adulterated with less expensive ingredients, the recognition of coffee species, the classification of plastic polymer samples, and the detection of ethanol level in beverages without the need to open the bottle.