Open AccessSensory Evaluation of Odor Approximation Using NMF with Kullback-Leibler Divergence and Itakura-Saito Divergence in Mass Spectrum Space
Author(s) -
Dani Prasetyawan,
Takamichi Nakamoto
Publication year - 2020
Publication title -
journal of the electrochemical society
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.258
H-Index - 271
eISSN - 1945-7111
pISSN - 0013-4651
DOI - 10.1149/1945-7111/abd110
Subject(s) - odor , divergence (linguistics) , non negative matrix factorization , kullback–leibler divergence , mathematics , set (abstract data type) , artificial intelligence , pattern recognition (psychology) , biological system , computer science , statistics , matrix decomposition , physics , biology , philosophy , linguistics , eigenvalues and eigenvectors , quantum mechanics , neuroscience , programming language
The odor reproduction can be achieved by approximating mass spectra of different odors by blending a set of odor components. The method enables us to create various odors by adjusting the blending recipe. The reproduced odor should be as close as possible to the target odor. Since there are no primary odors that have been found so far, finding an appropriate set of odor components to perform odor reproduction is essential. The number of odor components should be kept as small as possible whereas it should cover the widest range of odors. In the present study, we performed a sensory evaluation of odor reproduction. Odor reproduction and approximation by utilizing Nonnegative Matrix Factorization (NMF) particularly with Kullback-Leibler (KL) and Itakura-Saito (IS) divergences on mass spectrum space were evaluated. The sensory test reveals that the accuracy of odor approximation by IS divergence were higher than that of KL divergence. Moreover, the combination of NMF with IS divergence with that of KL divergence improved the accuracy.