Type | Advanced Lecture (5 ECTS) |
Lecturers | Dr. Jilles Vreeken and Prof. Dr. Tobias Marschall |
Teaching Assistants | Jonas Fischer (lead), Hufsah Ashraf (deputy), and Osman Ali Mian |
Tutors | Daniel Radke and Nancy Mekountchou Menoudjeu |
esl-ta (at) mpi-inf.mpg.de | |
Lectures |
Thursdays, 10–12 o'clock in Campus E2.1 (CBI), Room 0.01 |
Tutorials |
Mondays, 12–14 o'clock in E1.4 (MPII) Room 0.21, and in E2.1 (CBI) Room 0.01 Wednesdays, 12–14 o'clock in E1.4 (MPII) Room 0.21, and in E2.1 (CBI) Room 0.07 |
Office Hours |
Jilles Vreeken and
Tobias Marschall: after each lecture Jonas Fischer, Hufsah Ashraf, Osman Ali Mian: by appointment |
Summary |
In this course we will convey the ability, given a data set, to choose an appropriate statistical method for analyzing it, to select the appropriate parameters for the statistical model generated by that method and to assess the quality of the resulting model. Both theoretical and practical aspects will be covered. What we cover will be relevant for computer scientists in general as well as for other scientists involved in data analysis and modeling. |
Prerequisites |
The course is targeted to advanced students in computer science, bioinformatics, math, and general science with a mathematical background. Students should know linear algebra and have good basic knowledge of statistics. |
Month | Day | Topic | Slides | Assignment | Req. Reading | Opt. Reading |
---|---|---|---|---|---|---|
Oct | 17 | Introduction and Basics | 1st assignment out | ESL 1, 2, ISLR 1, 2 | ||
24 | Linear Regression I | ESL 3, ISLR 3 | ||||
31 | Linear Regression II | deadline 1st, 2nd out | ESL 3, ISLR 3 | |||
Nov | 7 | Classification I | ESL 4, ISLR 4 | |||
14 | Classification 2 | deadline 2nd, 3rd out | ESL 4, ISLR 4 | |||
21 | Resampling Methods | ESL 7, ISLR 5 | ||||
28 | Model Selection and Regularization | deadline 3rd, 4th out | ESL 3, ISLR 6 | |||
Dec | 5 | Dimensionality Reduction | ESL 3, ISLR 6 | |||
12 | Beyond Linear | deadline 4th, 5th out | ESL 5, 9, ISLR 7 | |||
19 | Trees and Forests | ESL 9, ISLR 8 | ||||
26 | yay holiday – no class | |||||
Jan | 2 | yay holiday – no class | deadline 5th, 6th out | |||
9 | Support Vector Machines | ESL 12, ISLR 9 | ||||
16 | Neural Networks | deadline 6th, 7th out | tbd | |||
23 | Unsupervised Learning | [3] | ||||
30 | Clustering | deadline 7th | ESL 14, ISLR 10 | |||
Feb | 6 | Association Discovery | tbd | |||
12 | registration deadline exam | |||||
19-21 | oral exams | |||||
March | 11-12 | re-exams |
The course will, by and large, follow the book "An Introduction to Statistical Learning with Applications in R" [1]. At times the course will take additional material from the book "The Elements of Statistical Learning" [2]. The former book is the more introductory text, the latter book is more advanced. Both books are available for as free PDFs. We strongly encourage you, though, to acquire at least the first book in print. Further background literature is available in the library in the so-called Semesteraparat.
[1] | An Introduction to Statistical Learning with Applications in R. Springer, 2013. |
[2] | The Elements of Statistical Learning. Springer, 2009. |
For selected lectures we will identify interesting optional reading, such as relevant recent research papers. These we will make available here.
[3] | Visualizing data using t-SNE. Journal of Machine Learning Research, 9:2579-2605, 2008. |
Each problem set will cover theoretical proofs and programming exercises with roughly equal weight. In general, the deadlines are on the day indicated in the schedule at 10:00 Saarbrücken standard-time. You are free to hand in earlier. Further details will be announced in the first lecture.
As programming language we will use R – a language for statistical computing. It is freely available for Windows, Linux and Mac. As a vectorized programming language, it is ideally suited for the problems we will encounter. There are also many freely available packages (or libraries) to perform a variety of classification and regression tasks, or to visualize the results of statistical analyses in a convenient way.
You hand in your solution as follows. For the theoretical exercises, you may hand in your solutions in handwritten form before the lecture, or send one PDF file with all the answers by email to esl-ta (at) mpi-inf.mpg.de. For the programming exercises, send a single email with both your R code as .R file (should compile with the command "Rscript YourCode.R") as well as a pdf answering the questions and showing the generated plots (if any).
No. | Handout | Date due | Discussed on | Assignment Sheet | Additional Material |
---|---|---|---|---|---|
1 | 17 Oct 2019 | 31 Oct 2019 | 5 and 7 Nov 2019 | Assignment 1 | data |
2 | 31 Oct 2019 | 14 Nov 2019 | 19 and 21 Nov 2019 | Assignment 2 | |
3 | 14 Nov 2019 | 28 Nov 2019 | 2 and 4 Dec 2019 | Assignment 3 | data |
4 | 28 Nov 2019 | 12 Dec 2019 | 16 and 18 Dec 2019 | available per the handout date |
There will be one tutorial per week. In the week after you submitted an assignment, the solution will be present in the tutorial sessions on Monday and Wednesday 12:00, repectively. We will also help you with the current problem set. In the following week, we will return the corrected sheets to you on Monday or Wednesday, respectively. We will also recapitulate the lectures, and have some time for discussions.
No. | Date | Slides |
---|---|---|
0 | 28/30 Oct 2019 | Tutorial 0, R-code, and Math foundations |
1 | 12/14 Nov 2019 | |
2 | 25/27 Nov 2019 | available per the day of the tutorial |
R (version 3.2.3) is installed on the CIP pool computers and can be started by invoking R
from the command line.
The official web site of the R project is r-project.org. You can download R for Windows, Linux and Mac from there. Additional packages, documentation and tutorials are also available for download from the official web site. Useful manuals and tutorials include:
The CRAN Contributed Documentation lists many other tutorials for R beginners and advanced programmers.
You can also check out RStudio, an open-source IDE for R.
You need a cumulative 50% of the points in the problem sets (in both theoretical and programming exercises) to be admitted to the exam.
To succesfully participate, you need to register for the exam in the LSF/HISPOS system of Saarland University – this will be possible as soon as the exam date has been entered into the system (this usually happens a few weeks into the semester).
The final exams will most likely be oral. The final decision on this will be made three weeks into the course. The final exam will cover all the material discussed in the lectures and the required reading. The main exam will be on February 19th, 20th, and 21st. The re-exam will be on March 11th, and 12th. The exact time slot per student will be announced per email. Inform the lecturers of any potential clashes as soon as you know them.
This course was originally developed by Thomas Lengauer, and we thank him for kindly providing his lecture materials and experience.