Modern Machine Learning (ML) algorithms are considered to be black-boxes we have no epis-temic access to. XAI tackles this assumption by providing methods that allow gaining insights into thebehaviour of ML algorithms. This course introduces the central philosophical concepts and challenges inXAI such as explanation, interpretability, and opacity. Moreover, we discuss state-of-the-art XAI meth-ods and their strengths and weaknesses. We focus particularly on causal explanations, the role of XAI for Science, and model-agnostic interpretation techniques.

The course is held together with students from MCMP.

Fridays 12:15-13:45 pm

Opening Event: Online

Key: xai-philo-21

Die empirischen Wissenschaften stehen vor wesentlichen Herausforderungen in Bezug auf die Reproduzierbarkeit und Replizierbarkeit ihrer Ergebnisse. Die sogenannte Replikationskrise beschreibt das Phänomen, dass eine Vielzahl an Studienergebnissen nicht auf unabhängigen Daten bestätigt werden können.

Dies liegt unter anderem an der Tatsache, dass in der Regel eine große Anzahl an verschiedenen Analysestrategien für eine bestimmte Forschungsfrage existiert (>>researcher degrees of freedom<<). Überdies werden die Ergebnisse oftmals nur selektiv für die gewählte Strategie veröffentlicht und die Variabilität in Bezug auf verschiedene andere Ansätze verschleiert (>>selective reporting<<).

Schwerpunkt des Seminars ist die Vermittlung von Grundlagen der Reproduzierbarkeit und Replikation in der Statistik. Die Studierenden werden dazu in die entsprechende Literatur und den Umgang mit geeigneten Tools, wie z.B. Git/Github (Versionskontrolle) und R Markdown (Reporting) eingeführt, um anschließend selbstständig an einer vorgegebenen Fragestellung (unter Einbezug der einschlägigen Literatur) zu arbeiten.

Vorbesprechung: Freitag, 22.10.2021 um 14:00-16:00 Uhr (s.t.) via Zoom
Einführungs-Kurs: Freitag 29.10.2021, 14:00-16:00 Uhr (s.t.) via Zoom
Ort: Das Seminar findet vsl. via Zoom statt.
Anrechnung: 6 ECTS oder 9 ECTS (6 ECTS (Seminar) + 3 ECTS (Wahlpflichtbereich))

Sie können sich in den Kurs mit dem Schlüssel DHS2122 selbst einschreiben.

Termine:

Die Veranstaltung "Statistische Software (R)" wendet sich an Studierende im Bachelorstudiengang Statistik (1. Semester).

- The enrollment key is preclin_clin_2122

Schedule 

• Lecture: Wednesday, 10 - 12 
• Lab session: Thursday, 12 - 14 
• Location: t.b.d.

Termine und Personen:

Description

The seminar deals with theoretical and practical concepts of uncertainty quantification in deep learning. 

Organization and Dates

Kick-off event: Friday, 15. October, 3 - 5 pm

Due to Corona, all talks will be online!

Enrolment Key

UQDL22

Target Audience 

• Statistics MSc. (Methods/Bio/WISO)
• Statistics and Data Science MSc.
  
• Data Science MSc.

Vorlesung und Übung: Frauke Kreuter, Julian Rodemann

Einschreibeschlüssel: WiSo202122

Der Einschreibeschlüssel lautet: stat3nf2022

Use the registration key (Einschreibeschlüssel) "bigDS" to read more about the course.
This course aims to foster the practice of software engineering and project management techniques in the context of data science and machine learning projects. 

The target audience for this course is master's students from Computer Science, Statistics & Data Science.

Organization

• Lecturers: Bernd Bischl, Giuseppe Casalicchio, Susanne Dandl,  Florian Pfisterer
• Time: During workshops: Thursday, 8-12 s.t (Only first weeks)
• Place: Online in Zoom
  
• ECTS: 12 ECTS (e.g. as an alternative to the Statistical Consulting or Data Science Practical)

Eligibility Requirements

• Good programming skills in Data science related languages (R, Python, Julia, C++, etc.)
• Predictive Modelling, KDD, Deep Learning or comparable Machine Learning courses

Projects

• Industry Projects with several Munich-based industry leaders
• Research / Data Science for Social Good projects 

Einschreibeschlüssel Selbsteinschreibung / Gast: APR2022

This course is closely related to the Generalized Regression Models (Generalisierte Regressions/GRM) course, also taught by Prof. Dr. Kuechenhoff. This course is, however, taught in English and designed for students currently attaining their master’s degree in statistics.

We will take a closer look at generalized models, mixed models, Bayesian approaches, generalized additive models, survival analysis and error models. This is not an exhaustive list, but should provide you with a general idea of the contents of the course.

If you have questions regarding the course, contact:

Kuechenhoff, Helmut <kuechenhoff@stat.uni-muenchen.de>

Rave, Martje <martje.rave@stat.uni-muenchen.de>

Lecture (Robert Czudaj)

Tuesday, 4 p.m. -  6 p.m., (Start: October 19, 2021)

Tutorial (Christoph Berninger)

Thursday, 6 p.m. -  8 p.m.,  (Start: October 28, 2021)

• Der Einschreibeschlüssel lautet: FinEco_2021_22

Termine

• Vorlesung: Dienstag, 12:00 - 14:00 Uhr,
• Übungen:  Montag, 14:00 - 16:00 Uhr

Einschreibeschlüssel

• Der Einschreibeschlüssel lautet "einf_oek_w2122".

Schedule

• The enrollment key is "stat_inf_ws2122"

This bachelor/master-seminar deals with selected methods and topics that are less common in the classical statistics curriculum .

Besides rather classical topics like non-parametric statistics, some non-classical topics aim at applications of order and lattice theory, others are rather motivated by the philosophy of science in the context of statistics (including machine learning contexts), further topics deepen aspects of decision theory.  For the latter, it is advantageous to have heard the corresponding lecture on decision theory, but this is not a must.

Since the seminar is offered for advanced bachelor students as well as master students and since the mathematical background of the students is probably relatively heterogeneous, this will be taken into account concerning the requirements and also in the supervision. In general, the exact design of the seminar topics (whether more in depth or more in breadth, etc.) can of course be individually shaped by the students in consultation with the lecturers.

The topics are partly related to each other and partly rather independent of each other.

The seminar will take place in blocks at the end of the semester (around the first week of March), the exact scheduling will take place in the preliminary meeting.

The preliminary meeting, during which the topic will be assigned, is tentatively scheduled for Monday, Nov. 15, beginning at 18:00 s.t. via zoom.

There will be an interim meeting (via zoom) around mid-January.

The seminar is planned as a virtual event via zoom.

Audience:

• Master Statistik: Räumliche Statistik
• Master Biostatistik: Räumliche Statistik
• Master Statistik WiSo: Räumliche Statistik
• Master Statistics and Data Science: Spatial Statistics
• Master Data Science ESG: Elective

Date/Time:

• Monday 10–12
• Wednesday 14–16

Language:

• The lecture will be in English, unless all students agree to German

Enrollment key:

• Enrollment key is "Moran".