Merge pull request #187 from chhoumann/svr-cite

SVR add cite

report_thesis/src/references.bib

@@ -617,21 +617,40 @@ @book{burkovHundredpageMachineLearning2023
   annotation = {OCLC: 1417057084}
 }
 
+@article{smola_svr_tutorial_2004,
+	title = {A tutorial on support vector regression},
+	volume = {14},
+	issn = {1573-1375},
+	url = {https://doi.org/10.1023/B:STCO.0000035301.49549.88},
+	doi = {10.1023/B:STCO.0000035301.49549.88},
+	abstract = {In this tutorial we give an overview of the basic ideas underlying Support Vector (SV) machines for function estimation. Furthermore, we include a summary of currently used algorithms for training SV machines, covering both the quadratic (or convex) programming part and advanced methods for dealing with large datasets. Finally, we mention some modifications and extensions that have been applied to the standard SV algorithm, and discuss the aspect of regularization from a SV perspective.},
+	language = {en},
+	number = {3},
+	urldate = {2024-06-04},
+	journal = {Statistics and Computing},
+	author = {Smola, Alex J. and Schölkopf, Bernhard},
+	month = aug,
+	year = {2004},
+	keywords = {machine learning, regression estimation, support vector machines},
+	pages = {199--222},
+}
+
 @article{shi_svr_libs_2015,
 	title = {Quantitative analysis of sedimentary rocks using laser-induced breakdown spectroscopy: comparison of support vector regression and partial least squares regression chemometric methods},
 	volume = {30},
 	issn = {0267-9477, 1364-5544},
 	shorttitle = {Quantitative analysis of sedimentary rocks using laser-induced breakdown spectroscopy},
 	url = {https://xlink.rsc.org/?DOI=C5JA00255A},
 	doi = {10.1039/C5JA00255A},
-  abstract = {Laser Induced Breakdown Spectroscopy (LIBS) is attracting more and more attention in geology fields because of its unique advantages of on-line and in situ analysis and the availability of portable even handheld instruments due to the development of laser sources and mini-spectrometers. However, parameters such as accuracy and precision of the instrument are still essential for field application. In this paper, two algorithms to determine the concentrations of five main elements (Si, Ca, Mg, Fe and Al) in sedimentary rock samples are proposed based on support vector regression (SVR) and partial least squares regression (PLSR). The proposed comparison demonstrates that the SVR model performed better with more satisfactory accuracy and precision under the optimized conditions. For SVR quantitative analysis, the spectral features (20 lines) without principal component analysis (PCA) were selected as the input variables. The optimized penalty parameter C and the key parameter of the radial basis function (RBF)-s obtained by genetic algorithm (GA) were 4.63 and 0.9159, respectively. Also, the best number of the principal components of PLSR was optimized to be 8 by 10-fold cross-validation (CV) testing. Furthermore, the accuracy corresponding to the average relative standard deviations (RSDs) and the precision related to the root mean square error (RMSE) were calculated according to the performance of the two regression models. A significant enhancement of accuracy, of up to 43.50 times, and of precision, of 7.19 times, for the SVR model was obtained, which can eliminate the self-absorption of plasma efficiently compared with the linear machine learning method PLSR. In conclusion, the chemometric method of SVR with better accuracy and precision can be successfully applied for the quantitative analysis of complex geological samples using the LIBS technique.}
+        abstract = {Laser Induced Breakdown Spectroscopy (LIBS) is attracting more and more attention in geology fields because of its unique advantages of on-line and in situ analysis and the availability of portable even handheld instruments due to the development of laser sources and mini-spectrometers. However, parameters such as accuracy and precision of the instrument are still essential for field application. In this paper, two algorithms to determine the concentrations of five main elements (Si, Ca, Mg, Fe and Al) in sedimentary rock samples are proposed based on support vector regression (SVR) and partial least squares regression (PLSR). The proposed comparison demonstrates that the SVR model performed better with more satisfactory accuracy and precision under the optimized conditions. For SVR quantitative analysis, the spectral features (20 lines) without principal component analysis (PCA) were selected as the input variables. The optimized penalty parameter C and the key parameter of the radial basis function (RBF)-s obtained by genetic algorithm (GA) were 4.63 and 0.9159, respectively. Also, the best number of the principal components of PLSR was optimized to be 8 by 10-fold cross-validation (CV) testing. Furthermore, the accuracy corresponding to the average relative standard deviations (RSDs) and the precision related to the root mean square error (RMSE) were calculated according to the performance of the two regression models. A significant enhancement of accuracy, of up to 43.50 times, and of precision, of 7.19 times, for the SVR model was obtained, which can eliminate the self-absorption of plasma efficiently compared with the linear machine learning method PLSR. In conclusion, the chemometric method of SVR with better accuracy and precision can be successfully applied for the quantitative analysis of complex geological samples using the LIBS technique.}
 	language = {en},
 	number = {12},
 	urldate = {2024-06-04},
 	journal = {Journal of Analytical Atomic Spectrometry},
 	author = {Shi, Qi and Niu, Guanghui and Lin, Qingyu and Xu, Tao and Li, Fengjun and Duan, Yixiang},
 	year = {2015},
 	pages = {2384--2393},
+}
 
 @article{el_haddad_ann_2013,
 	title = {Artificial neural network for on-site quantitative analysis of soils using laser induced breakdown spectroscopy},

report_thesis/src/sections/background/linear_and_regularization_models/svr.tex

@@ -12,7 +12,7 @@ \subsubsection{Support Vector Regression (SVR)}
 The instances right on or inside the margin are called \textit{support vectors}, which are used to 'support' the margin and decision boundary.
 
 \gls{svr} extends the principles of \gls{svm} to regression problems.
-We use our previous discussion of \gls{svm} to introduce \gls{svr} based on \citet{druckerSVR}.
+We use our previous discussion of \gls{svm} to introduce \gls{svr} based on \citet{druckerSVR} and \citet{smola_svr_tutorial_2004}.
 
 \gls{svr} aims to fit a function that predicts continuous values rather than finding the hyperplane that best separates data points.
 Instead of using a hyperplane to separate the data, \gls{svr} uses two parallel hyperplanes to define a margin within which the function should lie, often referred to as the $\epsilon$-\textit{tube}, where $\epsilon$ is a hyperparameter that defines the width of the tube.