Interpreting convolutional neural network classifiers applied
to laser-induced breakdown optical emission spectra

J 2024

Interpreting convolutional neural network classifiers applied to laser-induced breakdown optical emission spectra

KÉPEŠ, Erik; Jakub VRÁBEL; Tomáš BRÁZDIL; Petr HOLUB; Pavel POŘÍZKA et al.

Základní údaje

Originální název

Interpreting convolutional neural network classifiers applied to laser-induced breakdown optical emission spectra

Autoři

KÉPEŠ, Erik; Jakub VRÁBEL; Tomáš BRÁZDIL; Petr HOLUB; Pavel POŘÍZKA a Jozef KAISER

Vydání

Talanta, AMSTERDAM, Elsevier, 2024, 0039-9140

Další údaje

Jazyk

angličtina

Typ výsledku

Článek v odborném periodiku

Obor

10200 1.2 Computer and information sciences

Stát vydavatele

Nizozemské království

Utajení

není předmětem státního či obchodního tajemství

Odkazy

URL

Impakt faktor

Impact factor: 6.100

Označené pro přenos do RIV

Ano

Kód RIV

RIV/00216224:14330/24:00137458

Organizační jednotka

Fakulta informatiky

Klíčová slova anglicky

Laser-induced breakdown spectroscopy; Classification; Interpretable machine learning; Convolutional neural networks; ChemCam calibration dataset

Štítky

rivok

Příznaky

Mezinárodní význam, Recenzováno

Změněno: 4. 4. 2025 12:13, RNDr. Pavel Šmerk, Ph.D.

Anotace

V originále

Laser-induced breakdown spectroscopy (LIBS) is a well-established industrial tool with emerging relevance in high-stakes applications. To achieve its required analytical performance, LIBS is often coupled with advanced pattern-recognition algorithms, including machine learning models. Namely, artificial neural networks (ANNs) have recently become a frequently applied part of LIBS practitioners' toolkit. Nevertheless, ANNs are generally applied in spectroscopy as black-box models, without a real insight into their predictions. Here, we apply various post-hoc interpretation techniques with the aim of understanding the decision-making of convolutional neural networks. Namely, we find synthetic spectra that yield perfect expected classification predictions and denote these spectra class-specific prototype spectra. We investigate the simplest possible convolutional neural network (consisting of a single convolutional and fully connected layers) trained to classify the extended calibration dataset collected for the ChemCam laser-induced breakdown spectroscopy instrument of the Curiosity Mars rover. The trained convolutional neural network predominantly learned meaningful spectroscopic features which correspond to the elements comprising the major oxides found in the calibration targets. In addition, the discrete convolution operation with the learnt filters results in a crude baseline correction.

Citovat

KÉPEŠ, Erik; Jakub VRÁBEL; Tomáš BRÁZDIL; Petr HOLUB; Pavel POŘÍZKA a Jozef KAISER. Interpreting convolutional neural network classifiers applied to laser-induced breakdown optical emission spectra. Talanta. AMSTERDAM: Elsevier, 2024, roč. 266, č. 124946, s. 1-11. ISSN 0039-9140. Dostupné z: https://doi.org/10.1016/j.talanta.2023.124946.

@article{2447199,
   author = {Képeš, Erik and Vrábel, Jakub and Brázdil, Tomáš and Holub, Petr and Pořízka, Pavel and Kaiser, Jozef},
   article_location = {AMSTERDAM},
   article_number = {124946},
   doi = {https://doi.org/10.1016/j.talanta.2023.124946},
   keywords = {Laser-induced breakdown spectroscopy; Classification; Interpretable machine learning; Convolutional neural networks; ChemCam calibration dataset},
   language = {eng},
   issn = {0039-9140},
   journal = {Talanta},
   title = {Interpreting convolutional neural network classifiers applied to laser-induced breakdown optical emission spectra},
   url = {https://www.sciencedirect.com/science/article/pii/S0039914023006975?via%3Dihub},
   volume = {266},
   year = {2024}
}

TY  - JOUR
ID  - 2447199
AU  - Képeš, Erik - Vrábel, Jakub - Brázdil, Tomáš - Holub, Petr - Pořízka, Pavel - Kaiser, Jozef
PY  - 2024
TI  - Interpreting convolutional neural network classifiers applied to laser-induced breakdown optical emission spectra
JF  - Talanta
VL  - 266
IS  - 124946
SP  - 1-11
EP  - 1-11
PB  - Elsevier
SN  - 00399140
KW  - Laser-induced breakdown spectroscopy
KW  - Classification
KW  - Interpretable machine learning
KW  - Convolutional neural networks
KW  - ChemCam calibration dataset
UR  - https://www.sciencedirect.com/science/article/pii/S0039914023006975?via%3Dihub
N2  - Laser-induced breakdown spectroscopy (LIBS) is a well-established industrial tool with emerging relevance in high-stakes applications. To achieve its required analytical performance, LIBS is often coupled with advanced pattern-recognition algorithms, including machine learning models. Namely, artificial neural networks (ANNs) have recently become a frequently applied part of LIBS practitioners' toolkit. Nevertheless, ANNs are generally applied in spectroscopy as black-box models, without a real insight into their predictions. Here, we apply various post-hoc interpretation techniques with the aim of understanding the decision-making of convolutional neural networks. Namely, we find synthetic spectra that yield perfect expected classification predictions and denote these spectra class-specific prototype spectra. We investigate the simplest possible convolutional neural network (consisting of a single convolutional and fully connected layers) trained to classify the extended calibration dataset collected for the ChemCam laser-induced breakdown spectroscopy instrument of the Curiosity Mars rover. The trained convolutional neural network predominantly learned meaningful spectroscopic features which correspond to the elements comprising the major oxides found in the calibration targets. In addition, the discrete convolution operation with the learnt filters results in a crude baseline correction.
ER  -

KÉPEŠ, Erik; Jakub VRÁBEL; Tomáš BRÁZDIL; Petr HOLUB; Pavel POŘÍZKA a Jozef KAISER. Interpreting convolutional neural network classifiers applied to laser-induced breakdown optical emission spectra. \textit{Talanta}. AMSTERDAM: Elsevier, 2024, roč.~266, č.~124946, s.~1-11. ISSN~0039-9140. Dostupné z: https://doi.org/10.1016/j.talanta.2023.124946.

Přehled o publikaci