updates on titles, cites and l2 plots in regu

Signed-off-by: Ziyu-Mu <mu.ziyu.ovo@gmail.com>

slides/advriskmin/references.bib

@@ -0,0 +1,29 @@
+@inproceedings{SOLLICH1999NINTH,
+  author={Sollich, P.},
+  booktitle={1999 Ninth International Conference on Artificial Neural Networks ICANN 99. (Conf. Publ. No. 470)}, 
+  title={Probabilistic interpretations and Bayesian methods for support vector machines}, 
+  year={1999},
+  volume={1},
+  number={},
+  pages={91-96 vol.1},
+  keywords={},
+  doi={10.1049/cp:19991090},
+  url={https://ieeexplore.ieee.org/abstract/document/819547}
+}
+
+@inproceedings{MEYER2021ALTERNATIVE,
+  title={An alternative probabilistic interpretation of the huber loss},
+  author={Meyer, Gregory P},
+  booktitle={Proceedings of the ieee/cvf conference on computer vision and pattern recognition},
+  pages={5261--5269},
+  year={2021},
+  url={https://openaccess.thecvf.com/content/CVPR2021/papers/Meyer_An_Alternative_Probabilistic_Interpretation_of_the_Huber_Loss_CVPR_2021_paper.pdf}
+}
+
+@article{SALEH2022STATISTICAL,
+  title={Statistical properties of the log-cosh loss function used in machine learning},
+  author={Saleh, Resve A and Saleh, AK},
+  journal={arXiv preprint arXiv:2208.04564},
+  year={2022},
+  url={https://arxiv.org/pdf/2208.04564}
+}

slides/advriskmin/slides-advriskmin-bias-variance-decomposition-deepdive.tex

@@ -4,26 +4,24 @@
 \input{../../latex-math/basic-ml}
 \input{../../latex-math/ml-eval}
 
-\newcommand{\titlefigure}{figure/bias_variance_decomposition-linear_model_bias.png}
-\newcommand{\learninggoals}{
+\title{Introduction to Machine Learning}
+
+\begin{document}
+
+\titlemeta{% Chunk title (example: CART, Forests, Boosting, ...), can be empty
+    Advanced Risk Minimization
+  }{% Lecture title  
+    Bias-Variance Decomposition (Deep-Dive)
+  }{% Relative path to title page image: Can be empty but must not start with slides/
+  figure/bias_variance_decomposition-linear_model_bias.png
+  }{
   \item Understand how to decompose the generalization error of a learner into 
   \begin{itemize}
     \item \footnotesize Bias of the learner
     \item \footnotesize Variance of the learner
     \item \footnotesize Inherent noise in the data
   \end{itemize} 
-  }
-
-\title{Introduction to Machine Learning}
-\date{}
-
-\begin{document}
-
-\lecturechapter{Bias-Variance Decomposition (Deep-Dive)}
-\lecture{Introduction to Machine Learning}
-
-
-
+}
 
 \begin{vbframe} {Bias-Variance decomposition}
 

slides/advriskmin/slides-advriskmin-bias-variance-decomposition.tex

@@ -4,27 +4,25 @@
 \input{../../latex-math/basic-ml}
 \input{../../latex-math/ml-eval}
 
-\newcommand{\titlefigure}{figure/bias_variance_decomposition-linear_model_bias.png}
-\newcommand{\learninggoals}{
+\title{Introduction to Machine Learning}
+
+\begin{document}
+
+\titlemeta{% Chunk title (example: CART, Forests, Boosting, ...), can be empty
+    Advanced Risk Minimization
+  }{% Lecture title  
+    Advanced Risk Minimization:\\
+    Bias-Variance Decomposition
+  }{% Relative path to title page image: Can be empty but must not start with slides/
+  figure/bias_variance_decomposition-linear_model_bias.png
+  }{
   \item Understand how to decompose the generalization error of a learner into 
   \begin{itemize}
     \item \footnotesize bias of the learner
     \item \footnotesize variance of the learner
     \item \footnotesize inherent noise in the data
   \end{itemize} 
-  }
-
-\title{Introduction to Machine Learning}
-\date{}
-
-\begin{document}
-
-\lecturechapter{Advanced Risk Minimization:\\
-Bias-Variance Decomposition}
-\lecture{Introduction to Machine Learning}
-
-
-
+}
 
 \begin{vbframe} {Bias-Variance decomposition}
 

slides/advriskmin/slides-advriskmin-classification-01.tex

@@ -11,23 +11,20 @@
 \input{../../latex-math/basic-math}
 \input{../../latex-math/basic-ml}
 
-\newcommand{\titlefigure}{figure/plot_loss_01.png}
-\newcommand{\learninggoals}{
-  \item Derive the risk minimizer of the 0-1-loss
-  \item Derive the optimal constant model for the 0-1-loss
-}
-
 \title{Introduction to Machine Learning}
-% \author{Bernd Bischl, Christoph Molnar, Daniel Schalk, Fabian Scheipl}
-\institute{\href{https://compstat-lmu.github.io/lecture_i2ml/}{compstat-lmu.github.io/lecture\_i2ml}}
-\date{}
-
-
 
 \begin{document}
-
-\lecturechapter{0-1-Loss}
-\lecture{Introduction to Machine Learning}
+
+\titlemeta{% Chunk title (example: CART, Forests, Boosting, ...), can be empty
+    Advanced Risk Minimization
+  }{% Lecture title  
+    0-1-Loss
+  }{% Relative path to title page image: Can be empty but must not start with slides/
+  figure/plot_loss_01.png
+  }{
+  \item Derive the risk minimizer of the 0-1-loss
+  \item Derive the optimal constant model for the 0-1-loss
+}
 
 \begin{vbframe}{0-1-Loss}
 

slides/advriskmin/slides-advriskmin-classification-bernoulli.tex

@@ -14,28 +14,23 @@
 \input{../../latex-math/ml-eval}
 \input{../../latex-math/ml-trees} % For the comparison of Brier and Gini index
 
-\newcommand{\titlefigure}{figure/plot_bernoulli_prob}
-\newcommand{\learninggoals}{
+\title{Introduction to Machine Learning}
+
+\begin{document}
+
+\titlemeta{% Chunk title (example: CART, Forests, Boosting, ...), can be empty
+    Advanced Risk Minimization
+  }{% Lecture title  
+    Bernoulli Loss
+  }{% Relative path to title page image: Can be empty but must not start with slides/
+  figure/plot_bernoulli_prob.png
+  }{
   \item Know the Bernoulli loss and related losses (log-loss, logistic loss, Binomial loss)
   \item Derive the risk minimizer
   \item Derive the optimal constant model 
   \item Understand the connection between log-loss and entropy splitting 
 }
 
-\title{Introduction to Machine Learning}
-% \author{Bernd Bischl, Christoph Molnar, Daniel Schalk, Fabian Scheipl}
-\institute{\href{https://compstat-lmu.github.io/lecture_i2ml/}{compstat-lmu.github.io/lecture\_i2ml}}
-\date{}
-
-
-
-\begin{document}
-
-\lecturechapter{Bernoulli Loss}
-\lecture{Introduction to Machine Learning}
-
-
-
 \begin{vbframe}{Bernoulli Loss}
 
 \vspace*{-0.5cm}

slides/advriskmin/slides-advriskmin-classification-brier.tex

@@ -13,27 +13,23 @@
 \input{../../latex-math/basic-ml}
 \input{../../latex-math/ml-trees} % For the comparison of Brier and Gini index
 
+\title{Introduction to Machine Learning}
 
-\newcommand{\titlefigure}{figure/plot_brier.png}
-\newcommand{\learninggoals}{
+\begin{document}
+
+\titlemeta{% Chunk title (example: CART, Forests, Boosting, ...), can be empty
+    Advanced Risk Minimization
+  }{% Lecture title  
+    Brier Score
+  }{% Relative path to title page image: Can be empty but must not start with slides/
+  figure/plot_brier.png
+  }{
   \item Know the Brier score 
   \item Derive the risk minimizer
   \item Derive the optimal constant model 
   \item Understand the connection between Brier score and Gini splitting 
 }
 
-\title{Introduction to Machine Learning}
-% \author{Bernd Bischl, Christoph Molnar, Daniel Schalk, Fabian Scheipl}
-\institute{\href{https://compstat-lmu.github.io/lecture_i2ml/}{compstat-lmu.github.io/lecture\_i2ml}}
-\date{}
-
-
-\begin{document}
-
-\lecturechapter{Brier Score}
-\lecture{Introduction to Machine Learning}
-
-
 % \begin{vbframe}{Classification Losses: (Naive) L2-Loss}
 
 

slides/advriskmin/slides-advriskmin-classification-deepdive.tex

@@ -11,25 +11,22 @@
 \input{../../latex-math/basic-math}
 \input{../../latex-math/basic-ml}
 
-\newcommand{\titlefigure}
-{figure/plot_bernoulli_prob}
-\newcommand{\learninggoals}{
-  \item Derive the optimal constant model for the binary empirical log loss risk
-  \item Derive the optimal constant model for the empirical multiclass log loss risk
-}
+\newcommand{\argminl}{\mathop{\operatorname{arg\,min}}\limits}
 
 \title{Introduction to Machine Learning}
-% \author{Bernd Bischl, Christoph Molnar, Daniel Schalk, Fabian Scheipl}
-\institute{\href{https://compstat-lmu.github.io/lecture_i2ml/}{compstat-lmu.github.io/lecture\_i2ml}}
-\date{}
-
-\newcommand{\argminl}{\mathop{\operatorname{arg\,min}}\limits}
 
 \begin{document}
-
-\lecturechapter{Optimal constant model for the empirical log loss risk (Deep-Dive)}
-\lecture{Introduction to Machine Learning}
-
+
+\titlemeta{% Chunk title (example: CART, Forests, Boosting, ...), can be empty
+    Advanced Risk Minimization
+  }{% Lecture title  
+    Optimal constant model for the empirical log loss risk (Deep-Dive)
+  }{% Relative path to title page image: Can be empty but must not start with slides/
+  figure/plot_bernoulli_prob.png
+  }{
+  \item Derive the optimal constant model for the binary empirical log loss risk
+  \item Derive the optimal constant model for the empirical multiclass log loss risk
+}
 
 \begin{vbframe}{Binary log loss: Emp. Risk Minimizer}
 

slides/advriskmin/slides-advriskmin-classification-furtherlosses.tex

@@ -14,26 +14,23 @@
 \input{../../latex-math/ml-eval}
 \input{../../latex-math/ml-trees} % For the comparison of Brier and Gini index
 
+\title{Introduction to Machine Learning}
 
-\newcommand{\titlefigure}{figure/plot_loss_overview_classif.png}
-\newcommand{\learninggoals}{
+\begin{document}
+
+\titlemeta{% Chunk title (example: CART, Forests, Boosting, ...), can be empty
+    Advanced Risk Minimization
+  }{% Lecture title  
+    Advanced Classification Losses
+  }{% Relative path to title page image: Can be empty but must not start with slides/
+  figure/plot_loss_overview_classif.png
+  }{
   \item Know the (squared) hinge loss
   \item Know the $L2$ loss defined on scores
   \item Know the exponential loss
   \item Know the AUC loss 
 }
 
-\title{Introduction to Machine Learning}
-% \author{Bernd Bischl, Christoph Molnar, Daniel Schalk, Fabian Scheipl}
-\institute{\href{https://compstat-lmu.github.io/lecture_i2ml/}{compstat-lmu.github.io/lecture\_i2ml}}
-\date{}
-
-
-\begin{document}
-
-\lecturechapter{Advanced Classification Losses}
-\lecture{Introduction to Machine Learning}
-
 \begin{vbframe}{Hinge Loss}
 
 \begin{itemize}

slides/advriskmin/slides-advriskmin-logreg-deepdive.tex

@@ -11,26 +11,23 @@
 \input{../../latex-math/basic-math}
 \input{../../latex-math/basic-ml}
 
-\newcommand{\titlefigure}
-{figure/plot_bernoulli_prob}
-\newcommand{\learninggoals}{
-  \item Derive the gradient of the logistic regression
-  \item Derive the Hessian of the logistic regression
-  \item Show that the logistic regression is a convex problem
-}
+\newcommand{\argminl}{\mathop{\operatorname{arg\,min}}\limits}
 
 \title{Introduction to Machine Learning}
-% \author{Bernd Bischl, Christoph Molnar, Daniel Schalk, Fabian Scheipl}
-\institute{\href{https://compstat-lmu.github.io/lecture_i2ml/}{compstat-lmu.github.io/lecture\_i2ml}}
-\date{}
-
-\newcommand{\argminl}{\mathop{\operatorname{arg\,min}}\limits}
 
 \begin{document}
-
-\lecturechapter{Logistic regression (Deep-Dive)}
-\lecture{Introduction to Machine Learning}
-
+
+\titlemeta{% Chunk title (example: CART, Forests, Boosting, ...), can be empty
+    Advanced Risk Minimization
+  }{% Lecture title  
+    Logistic regression (Deep-Dive)
+  }{% Relative path to title page image: Can be empty but must not start with slides/
+  figure/plot_bernoulli_prob.png
+  }{
+  \item Derive the gradient of the logistic regression
+  \item Derive the Hessian of the logistic regression
+  \item Show that the logistic regression is a convex problem
+}
 
 \begin{vbframe}{Logistic regression: Risk Problem}
 

slides/advriskmin/slides-advriskmin-losses-properties.tex

@@ -15,29 +15,24 @@
 
 \usepackage{booktabs}
 
-\newcommand{\titlefigure}{figure_man/vgg_example.png}
-\newcommand{\learninggoals}{
+\title{Introduction to Machine Learning}
+
+\begin{document}
+
+\titlemeta{% Chunk title (example: CART, Forests, Boosting, ...), can be empty
+    Advanced Risk Minimization
+  }{% Lecture title  
+    Properties of Loss Functions
+  }{% Relative path to title page image: Can be empty but must not start with slides/
+  figure_man/vgg_example.png
+  }{
   % \item Understand why you should care about properties of loss functions
   \item Statistical properties 
   \item Robustness 
   \item Numerical properties
   \item Some fundamental terminology
-
 }
 
-\title{Introduction to Machine Learning}
-% \author{Bernd Bi{}schl, Christoph Molnar, Daniel Schalk, Fabian Scheipl}
-\institute{\href{https://compstat-lmu.github.io/lecture_i2ml/}{compstat-lmu.github.io/lecture\_i2ml}}
-\date{}
-
-
-\begin{document}
-
-% ------------------------------------------------------------------------------
-
-\lecturechapter{Properties of Loss Functions}
-\lecture{Introduction to Machine Learning}
-
 \begin{vbframe}{The role of Loss Functions}
 
 Why should we care about the choice of the loss function $\Lxy$?