Powered by OpenAIRE graph
Found an issue? Give us feedback
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/ ZENODOarrow_drop_down
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
ZENODO
Other literature type . 2025
License: CC BY
Data sources: ZENODO
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
ZENODO
Other literature type . 2025
License: CC BY
Data sources: ZENODO
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
ZENODO
Other literature type . 2025
License: CC BY
Data sources: ZENODO
ZENODO
Research . 2025
License: CC BY
Data sources: Datacite
ZENODO
Research . 2025
License: CC BY
Data sources: Datacite
ZENODO
Research . 2025
License: CC BY
Data sources: Datacite
ZENODO
Research . 2025
License: CC BY
Data sources: Datacite
versions View all 4 versions
addClaim

Come l'Autoencoder Convoluzionale Distingue gli Spettri Raman di Albite e Microclino (How a Convolutional Autoencoder Distinguishes Raman Spectra of Albite and Microcline)

Authors: Sparavigna, Amelia Carolina; Gemini (Modello Linguistico di Google);

Come l'Autoencoder Convoluzionale Distingue gli Spettri Raman di Albite e Microclino (How a Convolutional Autoencoder Distinguishes Raman Spectra of Albite and Microcline)

Abstract

Lo scopo del lavoro è testare l'efficacia di un autoencoder convoluzionale avanzato nella distinzione di minerali con spettri Raman estremamente simili. Albite e Microclino, due feldspati con spettri visivamente quasi identici, sono stati scelti per questa prova. Lo scopo quindi è determinare se l'autoencoder può identificare differenze sottili e non evidenti a un'analisi superficiale, producendo così pseudo-spettri distinti e rappresentativi per ciascun materiale. È stato creato un dataset dedicato utilizzando esclusivamente spettri di Albite e Microclino dal database RRUFF. Un aspetto cruciale è stata l'inclusione sia di spettri depolarizzati che polarizzati, per fornire all'autoencoder un set di dati più ricco e complesso con ulteriori indizi per la differenziazione. È stato impiegato un autoencoder convoluzionale come modello principale, poiché gli strati convoluzionali sono particolarmente efficaci nel riconoscere e astrarre schemi locali e complessi all'interno di un segnale, rendendolo ideale per l'analisi spettroscopica. L'autoencoder ha superato il compito in modo eccezionale. Dopo l'addestramento, che ha raggiunto un valore di perdita (loss) eccezionalmente basso, i dati sono stati analizzati con un algoritmo di clustering. Il risultato è stato una separazione quasi perfetta di Albite e Microclino in due cluster distinti. Sebbene un singolo spettro di Albite sia finito nel cluster di Microclino, il successo è notevole perché il modello ha trovato una distinzione chiara e robusta in un set di dati dove le differenze sono quasi impercettibili. Questo risultato dimostra che l'autoencoder non è solo uno strumento di denoising, ma un potente esploratore di dati, in grado di scoprire schemi e differenze nascoste all'interno di insiemi di dati complessi. La strategia di usare il modello per generare pseudo-spettri distinti è quindi pienamente convalidata. A conferma della generalizzabilità del metodo, un'appendice presenta i risultati di un'ulteriore applicazione su Calcite e Siderite, in cui è stata ottenuta una separazione perfetta. Aragonite e Calcite sono stati anche utilizzati mostrando che l’autoencoder è in grado di distinguere tali minerali. In appendice si discute anche come l’autoencoder lavori e come la scelta del clustering, nel nostro caso il K-means, possa decidere l’esito del risultato. The aim of this work is to test the effectiveness of an advanced convolutional autoencoder in distinguishing between minerals with extremely similar Raman spectra. Albite and Microcline, two feldspars with visually almost identical spectra, were chosen as the subjects for this stress test. The goal is to determine if the autoencoder could identify subtle, non-obvious differences, thus producing distinct and representative pseudo-spectra for each material. A dedicated dataset was created using only Albite and Microcline spectra from the RRUFF database. A crucial aspect of this step was the inclusion of both depolarized and polarized spectra. Using polarized data provided a richer and more complex feature set, offering the autoencoder additional clues for differentiation. A convolutional autoencoder was employed as the core model. This architecture was chosen because convolutional layers are highly effective at recognizing and abstracting local and complex patterns within a signal (such as the shape, width, and position of peaks), making it an ideal candidate for spectroscopic analysis. The autoencoder performed exceptionally well. After training, which resulted in an exceptionally low loss value, the data was analyzed using a clustering algorithm. The outcome was a near-perfect separation of Albite and Microcline into two distinct clusters. While one single Albite spectrum ended up in the Microcline cluster, the success is formidable because the model found a clear and robust distinction in a dataset where the differences are almost imperceptible. This result proves that the autoencoder is not just a denoising tool but a powerful data explorer capable of uncovering hidden patterns and differences within complex datasets. The strategy of using the model to generate distinct pseudo-spectra is thus fully validated. To confirm the generalizability of the method, an appendix presents the results of an additional application on Calcite and Siderite, where a perfect separation was achieved. Aragonite and Calcite have also been used showing that the autoencoder is able to distinguish such minerals. In the appendix we also discuss how the autoencoder works and how the choice of clustering, in our case the K-means, can decide the outcome of the result. 

Related Organizations
Keywords

Raman Spectroscopy, Microcline, Albite, Convolutional Autoencoder

  • BIP!
    Impact byBIP!
    selected citations
    These citations are derived from selected sources.
    This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    0
    popularity
    This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
    Average
    influence
    This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    Average
    impulse
    This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
    Average
Powered by OpenAIRE graph
Found an issue? Give us feedback
selected citations
These citations are derived from selected sources.
This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Citations provided by BIP!
popularity
This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
BIP!Popularity provided by BIP!
influence
This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Influence provided by BIP!
impulse
This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
BIP!Impulse provided by BIP!
0
Average
Average
Average
Green