Seminar: Robust Data Mining Techniques

News

• The session of Monday 24.07.17 will take place on Tuesday 25.07.17 at 13:00 in room 02.11.058. All other sessions will take place according to the regular schedule.
• Slides with organizational updates can be found here.

Description

Machine learning algorithms are getting a wide adoption across numerous domains of human activity. They are responsible for tasks ranging from content recommendation on the web to trading in the stock markets. At the same time, in many real-world scenarios the data contains imperfections that hinder the performance of these algorithms. For instance, in the industrial setting networks of sensors are prone to noise and random failures. On the internet, e-commerce platforms and social networks are subject to adversarial attacks by spammers and fraudsters. Such scenarios require novel data mining algorithms that are robust and immune to corruptions in the data.

The goal of the seminar is to familiarize the students with the state of the art in design of robust data mining algorithms. Topics discussed include both the extensions of classic machine learning algorithms aimed to increase robustness (e.g. PCA, spectral clustering), as well as high-level ideas surrounding the subject (e.g. differential privacy).

Topics

Date	Topic	Student	Supervisor	References	Reviewer 1	Reviewer 2
24.04	Robust Regression	Boonyakorn	Oleksandr	RANSAC Robust Regression Methods for Computer Vision: A Review *Peter J. Rousseeuw, Annick M. Leroy - Robust Regression and Outlier Detection	Daniela	Viet
08.05	Robust Classification	Nikolai	Amir	Learning with Noisy Labels Label-noise robust logistic regression and applications	Thomas	James
15.05	Robust Matrix Factorization	Maida	Aleksandar	Robust PCA Non-convex Robust PCA Robust Nonnegative Matrix Factorization via L1 Norm Regularization Robust Nonnegative Matrix Factorization Via Half-Quadratic Minimization Robust Nonnegative Matrix Factorization	Boonyakorn	Nikolai
22.05	Robust Clustering	Csongor	Roberto	Noise Robust Spectral Clustering Robust K-means: A Theoretical Revisit	Alexander	Viet
29.05	Robust Community Detection	Stevica	Aleksandar	On Community Outliers and their Efficient Detection in Information Networks Focused Clustering and Outlier Detection in Large Attributed Graphs Robust network community detection using balanced propagation	Nikolai	Csongor
12.06	Robust Time Series / Sequence Modeling	Lorenzo	Oleksandr	**Robust Statistics: Theory and Methods - Chapter 8 Learning an Outlier-Robust Kalman Filter Robust Multivariate Autoregression for Anomaly Detection in Dynamic Product Ratings	Thomas	Daniela
19.06	Attacks on Classifiers	Viet	Aleksandar	Poisoning Attacks against Support Vector Machines Adversarial Label Flips Attack on Support Vector Machines Evasion Attack of Multi-Class Linear Classifiers	Yuesong	Stevica
26.06	Fooling Deep Networks	James	Aleksandar	Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images DeepFool: A Simple and Accurate Method to Fool Deep Neural Networks A Theoretical Framework for Robustness of (Deep) Classifiers Under Adversarial Noise	Yuesong	Maida
03.07	Learning in the Adversarial Setting	Daniela	Aleksandar	Adversarial Classification Adversarial Support Vector Machine Learning Robustness of classifiers: from adversarial to random noise	Boonyakorn	Csongor
10.07	Learning from Crowds	Yuesong	Oleksandr	Learning from Crowds Supervised Learning from Multiple Experts: Whom to trust when everyone lies a bit	Alexander	Lorenzo
17.07	Differential Privacy	Thomas	Roberto	Differential Privacy The Algorithmic Foundations of Differential Privacy Signal Processing and Machine Learning with Differential Privacy	Maida	James
24.07	Robustness of Complex Networks	Alexander	Oleksandr	Network Robustness	Stevica	Lorenzo

* hardcopy of Peter J. Rousseeuw, Annick M. Leroy - Robust Regression and Outlier Detection is available in the TUM library.

Also available via Eaccess http://onlinelibrary.wiley.com.eaccess.ub.tum.de/book/10.1002/0471725382

** use Eaccess to access the PDFs of the chapters, i.e. onlinelibrary.wiley.com.eaccess.ub.tum.de/book/10.1002/0470010940

Organizational Details

• 12 Participants
• 5 ETCS
• Language: English
• Weekly meetings every Monday 14:30-16:00, room 02.09.14.
• Please send your questions regarding the seminar to kdd-seminar-robust(at)in.tum.de.

Prerequisites

• The seminar is intended for master students of the Computer Science department.
• This seminar deals with advanced and cutting edge topics in machine learning and data mining research. Therefore, the students are expected to have a solid background in these areas (e.g. having attended at least one of the related lectures, such as "Mining Massive Datasets", "Machine Learning", etc.). 

Requirements

• Extended abstract: 1 page article document class with motivation, key concepts and results.
• Paper: 5-8 pages in ACM format.
• Presentation: 30 minutes talk + 15 minutes discussion. (Optional: Beamer template)
• Peer-review process.
• Mandatory attendance of the weekly sessions.

Dates

• 27.01.2017 17:00: Pre-course meeting in Interims Hörsaal 2. Slides can be found here.
• 03.02.17 - 08.02.17: Application and registration in the matching system of the department
• After 15.02.17: Notification of participants
• 01.03.2017 11:00: Kick-off meeting in the room 02.09.014. Slides can be found here.
• Starting 24.04.17: Weekly meetings every Monday 14:30-16:00, room 02.09.14

Deadlines

• 1 week before the talk: submission of an extended abstract and slides
• One day before the talk: submission of a preliminary paper for review
• 1 week after the talk: receiving comments from reviewers
• 2 week after the talk: submission of the final paper