Understanding the patterns that neural networks learn from chemical spectra

Literature Information

Publication Date 2023-11-10
DOI 10.1039/D3DD00203A
Impact Factor 0
Authors

Laura Hannemose Rieger, Max Wilson, Tejs Vegge, Eibar Flores


View Original

Abstract

Analysing spectra from experimental characterization of materials is time consuming, susceptible to distortions in data, requires specific domain knowledge, and may be susceptible to biases in general heuristics under human analysis. Recent work has shown the potential of using neural networks to solve this task, and assist spectral interpretation with automated and unbiased analysis on-the-fly. However, the black-box nature of most neural networks poses challenges when interpreting which patterns from the data are being used to make predictions. Understanding how neural networks learn is critical to assess their accuracy on unseen data, justify critical decision-making based on predictions, and potentially unravel meaningful scientific insights. We present a 1D neural network to classify infrared spectra from small organic molecules according to their functional groups. Our model is within range of state-of-the-art performance while being significantly less complex than previously used networks reported in the literature. A smaller network reduces the risk of overfitting and enables exploring what the model has learned about the patterns in the spectra that relate to molecular structure and composition. With a novel two-step approach for explaining the neural network's classification process, our findings not only demonstrate that the model learns the characteristic group frequencies of functional groups, but also suggest it uses non-intuitive patterns such as tails and overtones when classifying spectra.

Related Literature

Identification of an emitting molecular species by time-resolved fluorescence applied to the excited state dynamics of pigment yellow 101

Seung Noh Lee, Jaeheung Park, Manho Lim, Taiha Joo

2014-03-26 Paper

DOI: 10.1039/C3CP54546F

Revisiting electroaccepting and electrodonating powers: proposals for local electrophilicity and local nucleophilicity descriptors

Christophe Morell, Alberto Vela, Frédéric Guégan, Henry Chermette

2014-10-03 Paper

DOI: 10.1039/C4CP03167A

Electronic structure investigation of the evanescent AtO+ ion

André Severo Pereira Gomes, Florent Réal, Nicolas Galland, Celestino Angeli, Renzo Cimiraglia, Valérie Vallet

2014-02-20 Paper

DOI: 10.1039/C3CP55294B

Shock wave and modeling study of the thermal decomposition reactions of pentafluoroethane and 2-H-heptafluoropropane

C. J. Cobos, L. Sölter, E. Tellbach

2013-12-18 Paper

DOI: 10.1039/C3CP54274B

Dimeric phenanthroimidazole for blue electroluminescent materials: the effect of substituted position attached to biphenyl center

Zhiming Wang, Ying Feng, Hui Li, Zhao Gao, Xiaojuan Zhang, Ping Lu, Ping Chen, Yuguang Ma, Shiyong Liu

2014-03-07 Paper

DOI: 10.1039/C4CP00209A

Orientation effects in morphology and electronic properties of anatase TiO2 one-dimensional nanostructures. II. Nanotubes

Dmitri B. Migas, Andrew B. Filonov, Victor E. Borisenko

2014-03-26 Paper

DOI: 10.1039/C3CP54906B

Electrodeposition of iron and iron–aluminium alloys in an ionic liquid and their magnetic properties

P. Giridhar, B. Weidenfeller, F. Endres

2014-03-26 Paper

DOI: 10.1039/C4CP00613E

Human serum albumin binding to silica nanoparticles – effect of protein fatty acid ligand

Joo Chuan Ang, Mark J. Henderson, Richard A. Campbell, Jhih-Min Lin, Peter N. Yaron, Andrew Nelson, Thomas Faunce, John W. White

2014-02-19 Paper

DOI: 10.1039/C4CP00293H

High DNP efficiency of TEMPONE radicals in liquid toluene at low concentrations

Nikolay Enkin, Guoquan Liu, Igor Tkach, Marina Bennati

2014-03-20 Communication

DOI: 10.1039/C4CP00854E

Electronic structure at nanocontacts of surface passivated CdSe nanorods with gold clusters

Deepashri Saraf, Anjali Kshirsagar

2014-02-27 Paper

DOI: 10.1039/C4CP00069B

You might also like

Compound Q&A

What are the main uses of 1H-Indazole-6-carbonitrile (CAS: 141290-59-7)?

1H-Indazole-6-carbonitrile finds applications in pharmaceuticals, where it serve...

141290-59-71H-Indazole-6-carbon...
Compound Q&A

How should waste containing Dioctyl (2E)-2-butenedioate (CAS: 2997-85-5) be handled?

Waste containing Dioctyl (2E)-2-butenedioate (CAS: 2997-85-5) should be collecte...

2997-85-5Dioctyl (2E)-2-buten...
Compound Q&A

What industries use Sodium [(1,2-benzoxazol-3-ylmethyl)sulfonyl]azanide (CAS: 68291-98-5)?

Sodium [(1,2-benzoxazol-3-ylmethyl)sulfonyl]azanide is primarily used in pharmac...

68291-98-5Sodium [(1,2-benzoxa...
Compound Q&A

Are there alternatives to Dimethyl 4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)-2,6-pyridinedicarboxylate (CAS: 741709-66-0) in synthesis?

Dimethyl 4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)-2,6-pyridinedicarboxyla...

741709-66-0Dimethyl 4-(4,4,5,5-...
Compound Q&A

How should waste containing 2-Fluoro-6-hydrazinopyridine (CAS: 80714-39-2) be handled?

Waste containing 2-Fluoro-6-hydrazinopyridine (CAS: 80714-39-2) should be manage...

80714-39-22-Fluoro-6-hydrazino...
Compound Q&A

What is 6-Formyl-2-pyridinecarboxylic acid (CAS: 499214-11-8)?

6-Formyl-2-pyridinecarboxylic acid is an organic compound with the molecular for...

499214-11-86-Formyl-2-pyridinec...
900874-91-13-(3,4-dimethoxyphen...
Compound Q&A

How is 9H-Tribenzo[b,d,f]azepine (CAS: 29875-73-8) typically synthesized?

9H-Tribenzo[b,d,f]azepine is typically synthesized via a multi-step process invo...

29875-73-89H-Tribenzo[b,d,f]az...
Compound Q&A

How is 1-Cyclopropyl-7-ethoxy-6-fluoro-8-methoxy-4-oxo-1,4-dihydro-3-quinolinecarboxylic acid (CAS: 1797982-51-4) typically synthesized?

1-Cyclopropyl-7-ethoxy-6-fluoro-8-methoxy-4-oxo-1,4-dihydro-3-quinolinecarboxyli...

1797982-51-41-Cyclopropyl-7-etho...
Compound Q&A

How should waste containing Methyl 3-oxo-1,2,3,4-tetrahydro-6-quinoxalinecarboxylate (CAS: 671820-52-3) be handled?

Waste containing Methyl 3-oxo-1,2,3,4-tetrahydro-6-quinoxalinecarboxylate (CAS: ...

671820-52-3Methyl 3-oxo-1,2,3,4...

Source Journal

Digital Discovery

Digital Discovery
CiteScore: 0
Self-citation Rate: 0%
Articles per Year: 0

Recommended Compounds

Recommended Suppliers

Disclaimer
This page provides academic journal information for reference and research purposes only. We are not affiliated with any journal publishers and do not handle publication submissions. For publication-related inquiries, please contact the respective journal publishers directly.
If you notice any inaccuracies in the information displayed, please contact us at support@chemtradehub.com. We will promptly review and address your concerns.