Deep learning modellering opent gesloten deuren in de chemie

Laurens
Van den Meersche

Denkt u dat een robot het ooit van ons overneemt? Zo ziet het er alvast niet naar uit! Een zelfdenkende robot die de mensheid ten onderbrengt in sciencefictionfilms, is immers niet waar artificiële intelligentie voor gemaakt is. Artificiële intelligentie moet het leven van de mens gemakkelijker maken en moet toelaten nieuwe werelden te ontdekken. Waar de menselijke intelligentie grenzen kent, kan de computer ons helpen deze te verleggen. Vanuit dat perspectief werd gehandeld in de masterthesis omtrent deep learning en chemisch kwantificeren. Er werd bewezen dat een computer wel degelijk de menselijke intelligentie kan verrijken in termen van een chemische kwantificatie.

De gezondheid van de planeet staat centraal

In de wereld waarin we vandaag leven, staat ecologie en milieuvriendelijkheid centraal. De chemische sector is niet bepaald de beste van de klas op dat gebied. Toch zijn er de laatste jaren veranderingen aan de gang.

Een chemisch product dat zo goed als in elk huishouden gebruikt wordt, zijn wasmiddelen. Vroeger bestonden deze uit synthetische chemicaliën die vaak schadelijk waren voor het milieu. De dagen van deze chemicaliën zijn geteld! Enzymen vervangen deze meer en meer. Ze zijn een hoogwaardig en milieuvriendelijk alternatief voor de schadelijke chemicaliën van vroeger. Door de toepassing van enzymen, zou er naar schatting vier keer minder energie en water nodig zijn om eenzelfde hoeveelheid wasgoed te wassen in vergelijking met vroeger.

Het draait niet alleen om enzymen

Enzymen zijn gevoelige moleculen en kunnen hun doeltreffendheid verliezen. Daarom moeten enzymen stabiel gehouden worden gedurende de periode tussen de productie en het moment dat ze in de wastrommel belanden. Daarvoor zijn stabilisatoren nodig. Dit zijn moleculen die ervoor zorgen dat het pas enzym werkzaam is op de moment dat het in contact komt met het wasgoed. Deze stabilisatoren zijn een essentieel onderdeel van het wasmiddel. Glycerine, monopropyleenglycol en sorbitol zijn voorbeelden van dergelijke stabilisatoren. Omdat deze zo belangrijk zijn, is er interesse om te weten hoeveel van die stabilisatoren er effectief aanwezig zijn in het wasmiddel. Dit is alleen gemakkelijker gezegd dan gedaan…

Wat is nu concreet het probleem?

De drie stabilisatoren die gebruikt worden, lijken qua chemische structuur sterk op elkaar. Hierdoor is het als mens moeilijk om een onderscheid te maken tussen deze moleculen. Er wordt daarom gekeken in de richting van de computer. De drie stabilisatoren worden bepaald via één infraroodspectrum. De computer blinkt uit in het analyseren van dergelijk spectrum. Hierdoor kan de computer beslissen welke piek toebehoort aan welke component en kan vervolgens door verwerking zeggen hoeveel van deze component er aanwezig is in een onbekend staal.

De computer bewees zijn kunnen

Hoewel er steeds van ‘de computer’ gesproken werd tot nog toe, is het juister om te spreken van een algoritme. Het gecreëerde algoritme bewees zijn kracht door naast het analyseren van een spectrum ook nog eens het spectrum te verbeteren. Bij de opname van een infraroodspectrum wordt namelijk gezien dat het spectrum naar boven verschuift. Door deze verschuiving wordt het nóg moeilijker om te kwantificeren. Het geschreven algoritme is erin geslaagd om dit spectrum te corrigeren door zelfstandig de basislijn van het spectrum terug naar beneden te verschuiven.

image-20210623163410-3

Kwantificatie van stabilisatoren, wat is dat?

Een kwantificatie van stabilisatoren is een bepaling van de aanwezige hoeveelheid stabilisatoren in een onbekend staal. Een onmogelijk opdracht voor het menselijk kunnen zo blijkt. Enerzijds omdat we het onderscheid niet kunnen maken tussen de stabilisatoren in een infraroodspectrum en anderzijds omdat we niet eeuwig de tijd hebben. Een analyse van een onbekend staal door een mens duurt ongeveer vijf minuten. Het geschreven algoritme analyseert ongeveer 500 stalen in minder dan één minuut.

Wat volgt er uit het onderzoek?

In het de masterthesis: Deep learning modellering voor de kwantitatieve FTIR-analyse van ternaire stabilisatorenmengsels, werd onderzocht of een computeralgoritme een kwantificatie van drie stabilisatoren tegelijkertijd kan uitvoeren met behulp van een infraroodspectrum en deep learning. Deep learning is een toepassing van artificiële intelligentie en wil zoveel zeggen als: “We laten de computer zelf zijn gang gaan en zelfstandig leren en verwerken”. Om dit zelfstandig leren mogelijk te maken, is een grote hoeveelheid aan stalen nodig.

Er werd aangetoond dat een computer tot op 0,56 % nauwkeurig een onbekend staal kan inschatten in de aanwezigheid van water. Dit is nauwkeuriger bepaald dan dat een mens dit ooit zal kunnen. In onderstaande figuur wordt de voorspelde concentratie uitgezet ten opzichte van de werkelijke waarde. Het model slaagt erin om, in de aanwezigheid van water, de stabilisatoren nauwkeurig te bepalen. De zwarte diagonaal stelt het ideale model voor, waarbij de voorspelde en de werkelijke concentratie aan elkaar gelijk zijn.

Het belang van deep learning

Zoals al aangehaald in de inleiding, moet artificiële intelligentie het leven gemakkelijker maken. In de masterthesis werd aangetoond dat dit inderdaad het geval is. Het denken van de computer zullen we nooit volledig kunnen volgen en dat kan bedreigend overkomen. Toch zijn er heel wat positieve kanten aan deep learning verbonden. Er werd in de thesis bewezen dat een computer nauwkeurig de concentraties kan bepalen van een stabilisatorenmengsel, maar dit kan evengoed doorgetrokken worden naar andere velden. Bijvoorbeeld het opsporen van zeldzame kankers kan ook gebeuren door middel van deep learning. We hebben nog maar het topje van de ijsberg gezien.

Waarom is het onderwerp van deze thesis nu zo belangrijk?

Als we verder willen evolueren naar een schoner milieu, moeten we in de chemische industrie meer in de richting van enzymen kijken. Deze belasten het milieu niet, in tegendeel. Om deze zo goed mogelijk in te zetten in de (wasmiddelen)industrie, moeten de enzymen stabiel gehouden worden waarvoor enzymstabilisatoren nodig zijn. De bepaling van de concentraties van de stabilisatoren is dan ook belangrijk en wordt mogelijk gemaakt door deep learning. Deze vorm van artificiële intelligentie geeft een extra dimensie aan de chemische wereld.

Artificiële intelligentie laat toe nieuwe werelden te ontdekken in de chemische industrie. Waar de menselijke intelligentie grenzen kent, helpt deep learning deze grenzen te verleggen en laat toe nieuwe inzichten te verweven. Het vooruitstrevend onderwerp van deze masterthesis, bewijst dat dit niet langer toekomstmuziek is.

 

Bibliografie

1.    Christeyns NV. Christeyns. [Internet]. 2021. [Geraadpleegd op 2021 Mar 2]. Via: https://www.christeyns.com/nl/professionele-textielverzorging.

2.    Olsen HS, Falholt P. The role of enzymes in modern detergency. J Surfactants Deterg. 1998; 1(4): 555–67.

3.    Infinita Biotech. What Are The Benefits Of Enzymes In Detergent? [Internet]. 2020. [Geraadpleegd op 2021 Mar 28]. Via: https://infinitabiotech.com/blog/benefits-of-enzymes-in-detergents/.

4.    Kahn Academy. Enzymes and the active site. [Internet]. 2021. [Geraadpleegd op 2021 Feb 10]. Via: https://www.khanacademy.org/science/ap-biology/cellular-energetics /enzyme-structure-and-catalysis/a/enzymes-and-the-active-site.

5.    Kravchenko GB. Enzyme Classifications . Kinetics Mechanisms of action Specificity and Regulation. [PowerPoint presentation]. 2012.

6.    Campbell PN. Enzymes in Industry: Chapter 3: General Production Methods. Vol. 6, Biochemical Education. 2004. 37–82 p.

7.    Niyonzima FN. Detergent-compatible bacterial cellulases. J Basic Microbiol. 2019; 59(2): 134–47.

8.    Kumari U, Singh R, Ray T, Rana S, Saha P, Malhotra K, et al. Validation of leaf enzymes in the detergent and textile industries: launching of a new platform technology. Plant Biotechnol J. 2019; 17(6): 1167–82.

9.    Shivanand P, Jayaraman G. Isolation and characterization of a metal ion-dependent alkaline protease from a halotolerant Bacillus aquimaris VITP4. Indian J Biochem Biophys. 2011; 48(2): 95–100.

10. Beck K. Different Types of Enzymes. [Internet]. Sciencing. 2018. [Geraadpleegd op 2021 Feb 10]. Via: https://sciencing.com/feedback-inhibition-important-regulating-enzyme-a….

11. Singh R, Kumar M, Mittal A, Mehta PK. Microbial enzymes: industrial progress in 21st century. 3 Biotech. 2016; 6(2): 1–15.

12. Vittaladevaram V. Fermentative Production of Microbial Enzymes and their Applications: Present status and future prospects. J Appl Biol Biotechnol. 2017; 5(04): 90–4.

13. Singh S, Singh J. Effect of polyols on the conformational stability and biological activity of a model protein lysozyme. AAPS PharmSciTech. 2003; 4(3): 1–9.

14. Castellanos IJ, Crespo R, Griebenow K. Poly(ethylene glycol) as stabilizer and emulsifying agent: A novel stabilization approach preventing aggregation and inactivation of proteins upon encapsulation in bioerodible polyester microspheres. J Control Release. 2003; 88(1): 135–45.

15. Mokhtar NF, Raja Noor Zaliha RNZR, Muhd Noor ND, Mohd Shariff F, Ali MSM. The immobilization of lipases on porous support by adsorption and hydrophobic interaction method. Catalysts. 2020; 10(7): 1–17.

16. Sipos B, Szilágyi M, Sebestyén Z, Perazzini R, Dienes D, Jakab E, et al. Mechanism of the positive effect of poly(ethylene glycol) addition in enzymatic hydrolysis of steam pretreated lignocelluloses. Comptes Rendus - Biol. 2011; 334(11): 812–23.

17. Iyer P V., Ananthanarayan L. Enzyme stability and stabilization-Aqueous and non-aqueous environment. Process Biochem. 2008; 43(10): 1019–32.

18. Plaxton WC. Avoiding Proteolysis during the Extraction and Purification of Active Plant Enzymes. Plant Cell Physiol. 2019; 60(4): 715–24.

19. Milosavić NB, Prodanović RM, Velićković D, Dimitrijević A. Macroporous poly(GMA-co-EGDMA) for enzyme stabilization. Vol. 1504, Methods in Molecular Biology. 2017. 139–147 p.

20. Braham SA, Siar EH, Arana-Peña S, Bavandi H, Carballares D, Morellon-Sterling R, et al. Positive effect of glycerol on the stability of immobilized enzymes: Is it a universal fact? Process Biochem. 2021; 102: 108–21.

21. Mienda BS, Yahya A, Galadima IA, Shamsir MS. An overview of microbial proteases for industrial applications. Res J Pharm Biol Chem Sci. 2014; 5(1): 388–96.

22. Bioninja. 7.6 Enzymes. [Internet]. 2016. [Geraadpleegd op 2021 Mar 27]. Via: http://www.old-ib.bioninja.com.au/higher-level/topic-7-nucleic-acids-an….

23. Sigma-Aldrich. PMSF: Phenylmethylsulfonyl fluoride. [Internet]. 2021. [Geraadpleegd op 2021 Mar 30]. Via: https://www.sigmaaldrich.com/catalog/product/roche/pmsfro ?lang=fr&region=BE.

24. Kumar SS, Sabu A. Quality Control and Downstream Processing of Therapeutic Enzymes. Vol. 1148, Advances in Experimental Medicine and Biology. 2019. 345–381 p.

25. Poku RA, Amissah F, Duverna R, Aguilar BJ, Kiros G-E, Lamango NS. Polyisoprenylated methylated protein methyl esterase as a putative drug target for androgen-insensitive prostate cancer. Ecancermedicalscience. 2014; 8: 459–459.

26. Jin J, Tarrant RD, Bolam EJ, Angell-Manning P, Soegaard M, Pattinson DJ, et al. Production, quality control, stability, and potency of cGMP-produced Plasmodium falciparum RH5.1 protein vaccine expressed in Drosophila S2 cells. npj Vaccines. 2018; 3(1).

27. Harris TK, Keshwani MM. Chapter 7: Measurement of Enzyme Activity. [Internet]. 1st ed. Vol. 463, Methods in Enzymology. Elsevier Inc.; 2009. 57–71 p. Via: http://dx.doi.org/10.1016/S0076-6879(09)63007-X.

28. Wang Y, Wang G, Moitessier N, Mittermaier AK. Enzyme Kinetics by Isothermal Titration Calorimetry: Allostery, Inhibition, and Dynamics. Front Mol Biosci. 2020; 7(October):                1–19.

29. Hadwan MH. Simple spectrophotometric assay for measuring catalase activity in biological tissues. BMC Biochem. 2018; 19(1): 1–8.

30. Hinton-Sheley P. What is ATR-FTIR? [Internet]. AZoLifeSciences. 2021. [Geraadpleegd op 2021 Feb 11]. Via: https://www.azolifesciences.com/article/What-is-ATR-FTIR.aspx.

31. Ausili A, Sánchez M, Gómez-Fernández JC. Attenuated total reflectance infrared spectroscopy: A powerful method for the simultaneous study of structure and spatial orientation of lipids and membrane proteins. Biomed Spectrosc Imaging. 2015; 4(2):                  159–70.

32. Paar A. Attenuated total reflectance (ATR). [Internet]. 2021. [Geraadpleegd op 2021 Mar 27]. Via: https://wiki.anton-paar.com/en/attenuated-total-reflectance-atr/#:~:tex… areas where the sample,total reflectance” (ATR).

33. Bradley M. Pathlength Considerations With ATR Sampling in FTIR. [Internet]. 2018. [Geraadpleegd op 2021 Mar 31]. Via: https://www.labcompare.com/10-Featured-Articles/352695-Pathlength-Consi….

34. Chemistry LibreTexts. Infrared Spectroscopy. [Internet]. 2020. [Geraadpleegd op 2021 Feb 11]. Via: https://chem.libretexts.org/Bookshelves/Physical_and_Theoretical_ Chemistry_Textbook_Maps/Supplemental_Modules_(Physical_and_Theoretical_Chemistry)/Spectroscopy/Vibrational_Spectroscopy/Infrared_Spectroscopy/Infrared_Spectroscpy#:~:text=Absorption of IR r.

35. Schelden V. FT-IR studie van afbraakprocessen van vluchtige organische stoffen via atmosferische plasma’s. UGent. UGent; 2008.

36. Williams M. What is the Speed of Light? [Internet]. 2016. [Geraadpleegd op 2021 Feb 15]. Via: https://www.universetoday.com/38040/speed-of-light-2/.

37. Stein J. Planck’s Constant: The Number That Rules Technology, Reality, and Life. [Internet]. The nature of reality. 2011. [Geraadpleegd op 2021 Feb 15]. Via: https://www.pbs.org/wgbh/nova/article/plancks-constant/.

38. Van de Voorde I. Spectroscopische technieken: infrarood. 2019th–2020th ed. Gent; 2019.

39. Van de Ven MFC, Qiu J, Zhang Y. Infrarood metingen op bitumen: Voorbeeld van mogelijkheden . 2012.

40. Winter A. How to Find Functional Groups in the IR Spectrum. [Internet]. 2021. [Geraadpleegd op 2021 Feb 12]. Via: https://www.dummies.com/education/science/ chemistry/how-to-find-functional-groups-in-the-ir-spectrum/.

41. Ed Vitz, John W. Moore, Justin Shorb, Xavier Prat-Resina, Tim Wendorff & AH. 21.5: The Spectra of Molecules- Infrared. [Internet]. 2020. [Geraadpleegd op 2021 Feb 11]. Via: https://chem.libretexts.org/Bookshelves/General_Chemistry/Book%3A_ChemP….

42. MacKie DM, Jahnke JP, Benyamin MS, Sumner JJ. Simple, fast, and accurate methodology for quantitative analysis using Fourier transform infrared spectroscopy, with bio-hybrid fuel cell examples. MethodsX. [Internet]. 2016 ;3: 128–38. Via: http://dx. doi.org/10.1016/j.mex.2016.02.002.

43. Klein O, Roth A, Dornuf F, Schöller O, Mäntele W. The good vibrations of beer. The use of infrared and UV/Vis spectroscopy and chemometry for the quantitative analysis of beverages. Zeitschrift fur Naturforsch - Sect B J Chem Sci. 2012; 67(10): 1005–15.

44. ThermoFischer. FTIR FAQs: Using the Beer-Lambert law in FT-IR ATR for quantitative analysis of a time-sensitive, migrating substance (e.g., erucamide) in a polymer is difficult. How can this be overcome? [Internet]. 2020. [Geraadpleegd op 2021 Mar 31]. Via: https://www.thermofisher.com/be/en/home/industrial/spectroscopy-element… alysis/spectroscopy-elemental-isotope-analysis-learning-center/molecular-spectroscopy-information/ftir-information/ftir-faqs.html.

45. Jiménez-Carvelo AM, Osorio MT, Koidis A, González-Casado A, Cuadros-Rodríguez L. Chemometric classification and quantification of olive oil in blends with any edible vegetable oils using FTIR-ATR and Raman spectroscopy. LWT - Food Sci Technol. 2017; 86: 174–84.

46. Mallah MA, Sherazi STH, Bhanger MI, Mahesar SA, Bajeer MA. A rapid Fourier-transform infrared (FTIR) spectroscopic method for direct quantification of paracetamol content in solid pharmaceutical formulations. Spectrochim Acta - Part A Mol Biomol Spectrosc [Internet]. 2015; 141 :64–70. Via: http://dx.doi.org/10.1016/j.saa.2015.01.036.

47. Dreissig I, Machill S, Salzer R, Krafft C. Quantification of brain lipids by FTIR spectroscopy and partial least squares regression. Spectrochim Acta - Part A Mol Biomol Spectrosc. 2009; 71(5): 2069–75.

48. Luo J, Ying K, Bai J. Savitzky-Golay smoothing and differentiation filter for even number data. Signal Processing. 2005; 85(7): 1429–34.

49. Alrezj OA, Patchava K, Benaissa M, Alshebeili SA. Pre-processing to Enhance the Quantitative Analysis of Glucose from NIR and MIR Spectra. [Internet]. Vol. 65, IFMBE Proceedings. 2018. Via: http://link.springer.com/10.1007/978-981-10-5122-7.

50. Crochet RB. Data processing data processing flowchart. 2017; 1–31.

51. Sigma-Aldrich. D-Sorbitol. [Internet]. 2021. [Geraadpleegd op 2021 Feb 12]. Via: https://www.sigmaaldrich.com/catalog/ product/sigma/S1876?lang=fr&region=BE.

52. TGI Chemicals. Glycerol. [Internet]. 2021. [Geraadpleegd op 2021 Feb 12]. Via: https://www.tcichemicals.com/BE/en/p/G0316v.

53. Danish M, Mumtaz MW, Fakhar M, Rashid U. Response surface methodology based optimized purification of the residual glycerol from biodiesel production process. Chiang Mai J Sci. 2017; 44(4): 1570–82.

54. Sigma-Aldrich. 1,2-Propanediol. [Internet]. 2021. [Geraadpleegd op 2021 Feb 12]. Via: https://www.sigmaaldrich.com/catalog/product/sigma/p1009?lang=fr&region….

55. He S, Zhang W, Lijuan L, Yu H, Jiming H, Wanyi X, et al. Baseline correction for Raman spectra using an improved asymmetric least squares method. Anal Methods. 2014;(6): 4402–7.

56. Eilers PHC. A perfect smoother. Anal Chem. 2003; 75(14): 3631–6.

57. He S, Fang S, Liu X, Zhang W, Xie W, Zhang H, et al. Investigation of a genetic algorithm based cubic spline smoothing for baseline correction of Raman spectra. Chemom Intell Lab Syst. [Internet]. 2016; 152: 1–9. Via: http://dx.doi.org/10.1016/j.chemolab. 2016.01.005.

58. Shao L, Griffiths PR. Automatic baseline correction by wavelet transform for quantitative open-path fourier transform infrared spectroscopy. Environ Sci Technol. 2007; 41(20): 7054–9.

59. Zhang F, Tang X, Tong A, Wang B, Wang J. An automatic baseline correction method based on the penalized least squares method. Sensors (Switzerland). 2020; 20(7).

60. Zhang ZM, Chen S, Liang YZ. Baseline correction using adaptive iteratively reweighted penalized least squares. Analyst. 2010; 135(5): 1138–46.

61. Héberger K. Chemoinformatics — multivariate mathematical – statistical methods. Med Appl mass Spectrom. 2008; 141–69.

62. Grisanti E, Totska M, Huber S, Calderon CK, Hohmann M, Lingenfelser D, et al. Dynamic localized SNV, Peak SNV, and partial peak SNV: Novel standardization methods for preprocessing of spectroscopic data used in predictive modeling. J Spectrosc. 2018; 2018.

63. Syvilay D, Wilkie-Chancellier N, Trichereau B, Texier A, Martinez L, Serfaty S, et al. Evaluation of the standard normal variate method for Laser-Induced Breakdown Spectroscopy data treatment applied to the discrimination of painting layers. Spectrochim Acta - Part B At Spectrosc. [Internet]. 2015; 114: 38–45. Via: http://dx.doi.org /10.1016/j.sab.2015.09.022.

64. Bock T. What is a Latent Variable. [Internet]. 2020. DisplayR. [Geraadpleegd op 2021 Feb 22]. Via: https://www.displayr.com/what-is-a-latent-variable/.

65. Saikat M, Jun Y, Maitra S, Yan J. Principle Component Analysis and Partial Least Squares : Two Dimension Reduction Techniques for Reegression. Casualty Actuar Soc. 2008; 79–90.

66. Javatpoint. Purpose of Normalization. [Internet]. Javatpoint. 2018. [Geraadpleegd op 2021 May 31]. Via: https://www.javatpoint.com/dbms-purpose-of-normalization.

67. Jaadie Z. WHEN AND WHY TO STANDARDIZE YOUR DATA? A simple guide on when it is necessary to standardize your data. Builtin. 2021.

68. Géron A. Hands-on machine learning with Scikit-Learn and TensorFlow : concepts, tools, and techniques to build intelligent systems. O’Reilly Media. 2017.

69. Chollet F. Deep Learning with Python. MITP-Verlags GmbH & Co. KG. Manning Publications Co.; 2018. 373 p.

70. Missing Link AI. Deep Learning for Big Data. [Internet]. 2020. [Geraadpleegd op 2021 Feb 20]. Via: https://missinglink.ai/guides/neural-network-concepts/deep-learning-for….

71. Frankenfield J. Artificial intelligence AI. [Internet]. 2021. [Geraadpleegd op 2021 Feb 15]. Via: https://www.investopedia.com/terms/a/artificial-intelligence-ai.asp.

72. Sperling E. Deep Learning Spreads. [Internet]. 2018. [Geraadpleegd op 2021 Feb 15]. Via: https://semiengineering.com/deep-learning-spreads/.

73. AISOLAB. INTRO TO AI #1: ARTIFICIAL INTELLIGENCE, MACHINE LEARNING, NEURAL NETWORKS AND DEEP LEARNING: WHAT ARE THE DIFFERENCES? [Internet]. 2017. [Geraadpleegd op 2021 Feb 20]. Via: https://aiso-lab.com/intro-to-ai-1-artificial-intelligence-machine-lear….

74. Technology. Differences between machine learning and software engineering. [Internet]. 2018. [Geraadpleegd op 2021 Feb 15]. Via: https://futurice.com/blog/differences-between-machine-learning-and-soft….

75. Grossfield B. Deep learning vs machine learning: a simple way to understand the difference. [Internet]. 2020. [Geraadpleegd op 2021 Feb 15]. Via: https://www.zendesk.com/blog/machine-learning-and-deep-learning/#:~:tex… recap the differences between,intelligent decisions on its own.

76. IBM Cloud Education. Neural networks. [Internet]. 2020. [Geraadpleegd op 2021 Feb 15]. Via: https://www.ibm.com/cloud/learn/neural-networks.

77. Rashid M, Singh H, Goyal V. The use of machine learning and deep learning algorithms in functional magnetic resonance imaging—A systematic review. Expert Syst. 2020; 37(6): 1–29.

78. Harikrishna B. Deep Learning. [Internet]. 2018. [Geraadpleegd op 2021 Feb 16]. Via: https://medium.com/datadriveninvestor/deep-learning-2025e8c4a50.

79. Amini A. 6S191_MIT_DeepLearning_L1.pdf. [PowerPoint presentation]. MIT; 2020.

80. Github Pages. Deep Learning From Scratch. [Internet]. Zied HY’s Data Science Blog. 2021. [Geraadpleegd op 2021 Feb 16]. Via: https://ziedhy.github.io/Introduction_ Deep_Learning.html.

81. Homenick C. What’s The Role Of Weights And Bias In a Neural Network? [Internet]. 2020. [Geraadpleegd op 2021 Feb 18]. Via: https://morioh.com/p/eb23d31bb742.

82. Feng J, He X, Teng Q. Commonly-used-activation-functions-a-Sigmoid-b-Tanh-c-ReLU-and-d-LReLU. [Internet]. 2019. [Geraadpleegd op 2021 Mar 30]. Via: https://www.researchgate.net/figure/Commonly-used-activation-functions-….

83. Doshi S. Various Optimization Algorithms For Training Neural Network. [Internet]. Towards data science. 2019. [Geraadpleegd op 2021 Mar 2]. Via: https: //towardsdatascience.com/optimizers-for-training-neural-network-59450d71caf6.

84. Vieira S, Lopez Pinaya WH, Garcia-Dias R, Mechelli A. Deep neural networks. [Internet]. Machine Learning: Methods and Applications to Brain Disorders. Elsevier Inc.; 2019. 157–172 p. Via: http://dx.doi.org/10.1016/B978-0-12-815739-8.00009-2.

85. Subir V, Sanjiv D. Chapter 7 Training Neural Networks Part 1. [Internet]. 2018. [Geraadpleegd op 2021 Feb 18]. Via: https://srdas.github.io/DLBook/.

86. Subir V, Sanjiv D. Chapter 8 Training Neural Networks Part 2. [Internet]. 2020. [Geraadpleegd op 2021 Feb 18]. Via: https://srdas.github.io/DLBook/ ImprovingModelGeneralization.html#detecting-underfitting.

87. Edspresso Team. Overfitting and underfitting. [Internet]. Edspresso. 2021. [Geraadpleegd op 2021 Feb 18]. Via: https://www.educative.io/edpresso/overfitting-and-underfitting.

88. Varma S, Sanjiv D. Chapter 9 Training Neural Networks Part 3. [Internet]. 2018. [Geraadpleegd op 2021 Feb 18]. Via: https://srdas.github.io/DLBook /HyperParameterSelection.html.

89. Amini A. 6S191_MIT_DeepLearning_L3.pdf. [PowerPoint presentation]. MIT; 2020.

90. Bharath R, Reza BZ. Chapter 4. Fully Connected Deep Networks. [Internet]. 2021. [Geraadpleegd op 2021 Feb 18]. Via: https://www.oreilly.com/library/view/tensorflow-for-deep/9781491980446/….

91. Amini A. 6S191_MIT_DeepLearning_L2.pdf. [PowerPoint presentation]. MIT; 2020.

92. NumPy. NumPy. [Internet]. 2020. [Geraadpleegd op 2021 Mar 23]. Via: https://numpy.org/.

93. Matiplot. Matplotlib: Visualization with Python. [Internet]. 2021. [Geraadpleegd op 2021 Mar 22]. Via: https://matplotlib.org/.

94. TensorFlow. TensorFlow Core. [Internet]. 2021. [Geraadpleegd op 2021 Mar 23]. Via: https://www.tensorflow.org/guide/tensor.

95. Pydata. Pandas: Package overvieuw. [Internet]. 2021. [Geraadpleegd op 2021 Mar 23]. Via: https://pandas.pydata.org/pandas-docs/stable/getting_started/overview.h….

96. Peng J, Peng S, Jiang A, Wei J, Li C, Tan J. Asymmetric least squares for multiple spectra baseline correction. Anal Chim Acta. [Internet]. 2010; 683(1): 63–8. Via: http://dx.doi. org/10.1016/j.aca.2010.08.033.

97. Github Gist: Perimosocordiae. Asymmetric Least Squares. [Internet]. 2015. [Geraadpleegd op 2021 Feb 20]. Via: https://gist.github.com/perimosocordiae/ efabc30c4b2c9afd8a83.

98. Liu X, Zhang Z, Sousa PFM, Chen C, Ouyang M, Wei Y, et al. Selective iteratively reweighted quantile regression for baseline correction. Anal Bioanal Chem. 2014; 406(7): 1985–98.

99. Github: zmzhang. AirPLS. [Internet]. github. 2016. [Geraadpleegd op 2021 Feb 20]. Via: https://github.com/zmzhang/airPLS/blob/master/airPLS.py.

100.  Pelliccia D. Two scatter correction techniques for NIR spectroscopy in Python. [Internet]. 2018. [Geraadpleegd op 2021 Feb 20]. Via: https://nirpyresearch.com/two-scatter-correction-techniques-nir-spectro….

101.  Galarnyk M. PCA using Python (scikit-learn). [Internet]. 2017. [Geraadpleegd op 2021 Mar 1]. Via: https://towardsdatascience.com/pca-using-python-scikit-learn-e653f8989e 60.

102.  Sealens M. Interpretatie en modellering van multi instrumentele analytische data met Deep Learning. KU Leuven; 2020.

103.  Halvorsen KB. Relationship between SVD and PCA. How to use SVD to perform PCA? [Internet]. StackExchange. 2015. [Geraadpleegd op 2021 Apr 24]. Via: https:// stats.stackexchange.com/questions/134282/relationship-between-svd-and-pca-how-to-use-svd-to-perform-pca.

104.  Beleites C. PCA and the train/test split. [Internet]. StackExchange. 2021. [Geraadpleegd op 2021 Apr 23]. Via: https://stats.stackexchange.com/questions/ 55718/pca-and-the-train-test-split.

105.  Uhrovčík J. Strategy for determination of LOD and LOQ values - Some basic aspects. Talanta. 2014; 119: 178–80.

Download scriptie (3.34 MB)
Universiteit of Hogeschool
KU Leuven
Thesis jaar
2021
Promotor(en)
Prof. Dr. Eddy Courtijn, Ing. Simon Lambert, Ing. Delphine Verstraete en Ing. Marjolein Saelens