XP36RGM -- Reading group in data mining and machine learning

Data mining aims at revealing non-trivial, hidden and ultimately applicable knowledge in large data. Data size and data heterogeneity make two key data mining technical issues to be solved. The main goal is to understand the patterns that drive the processes generating the data. Machine learning focuses at computer algorithms that can improve automatically through experience and by the use of data. It often puts emphasis on performance that the algorithms reach. The distinction between DM and ML is not strict as machine learning is often used as a means of conducting useful data mining. For this reason, we cover both the areas in the same course.

The course will take a form of reading and discussion group. Each student gives two 1 hour lectures, followed by a 30 min discussion. One of the lectures should be general (MMDS book chapters, recent tutorials at major ML/DM conferences, etc.), the second one can present your research (if ML/DM related) or a ML/DM topic that is closely related to your research or research interests. Each student is supposed to read a review paper recommended for the topic before presentations of the other students. It is assumed that students have completed at least some of the master courses on Machine Learning and Data Analysis (B4M36SAN, B4M46SMU, BE4M33SSU).

Go beyond the literature, provide own insight, offer own illustrative examples, etc.

The students who do not present are supposed to read recommended reading and prepare a couple of questions before the class. The questions will be discussed during or after the talk.

Fall 2021

L	Date	Presents	Contents	Reading	Talk, other links
1	Oct 1	JK, FZ	Course overview, introduction, research interests		Course overview
2	Oct 15	Xuzhe Dang	Model-based reinforcement learning	MuZero	Talk
3	Oct 22	Peter Jung	How neural networks changed the landscape of graph processing	GNNs	Talk
4	Nov 5	Herbert Ullrich	Automated Fact-Checking	AFC	Talk
5	Nov 5	Jaroslav Moravec	Outlier detection methods – RANSAC	USAC	Talk
6	Nov 12	Michaela Urbanovska	Neural Algorithmic Reasoning	NAR	Talk
7	Nov 19	Ondrej Lukas	Explainable AI	ExpNN	Talk
8	Nov 26	Lukas Korel	ML with ontologies	MLwO	Talk
9	Dec 3	Peter Jung	Generative Adversarial Networks	GANs	Talk
10	Dec 10	Xuzhe Dang	StyleGANs	StyleGANs	Talk
11	Dec 17	Herbert Ullrich	BERT-like encoder models	BERT	TempTalk,Demo
12	Jan 7	Ondrej Lukas	Federated learning	FederatedAveraging	Talk
13	Jan 14	Jaroslav Moravec	Visual space odometry	VSO	TempTalk
14	Jan 21	Michaela Urbanovska	Deep learning for automated planning	Geffner	Talk
15	Jan 28	Lukas Korel	Video Scene Location Recognition	ISR
x	Jan 28	JK, FZ	exam – see the plan in the table below

Final presentations = exam

Each participant prepares a 5-7min talk that summarizes the main ideas presented before by another course participant. The topic assignment is:

Topic	Date of the original presentation	For exam presented by
Model-based reinforcement learning	Oct 15	Lukas Korel
How neural networks changed the landscape of graph processing	Oct 22	Herbert Ullrich
Outlier detection methods – RANSAC	Nov 5	Xuzhe Dang
Neural Algorithmic Reasoning	Nov 12	Peter Jung
Explainable AI	Nov 19	Jaroslav Moravec
Generative Adversarial Networks	Dec 3 (Dec 10 could be used too)	Michaela Urbanovska
BERT-like encoder models	Dec 17	Ondrej Lukas

References

Recent papers: Distill papers, Optimization Methods for Large-Scale Machine Learning, Wasserstein GAN, XGBoost: A Scalable Tree Boosting System,Deep Forest, Quantum Machine Learning,
Rajaraman, A., Leskovec, J., Ullman, J. D.: Mining of Massive Datasets, Cambridge University Press, 2011.
Free Data mining Books
Recent tutorials, major ML/DM conferences:
- ICML: 2021, 2020, 2019,
- KDD: 2021, 2020,2019,
- ECML/PKDD: 2021, 2020, 2019,
- NeurIPS: 2021, 2020, 2019,
Review papers: Yang, Wu: 10 CHALLENGING PROBLEMS IN DATA MINING RESEARCH, Wu et al.: Top 10 algorithms in data mining
External seminars: ML seminars at MFF, PIS, Machine Learning Meetups, FIS KEG.

Links

Lecturers: Jiří Kléma, Filip Železný
Class schedule
Course syllabus (the former XP36VPD).

Evaluation, requirements

every student must give his talks (the principle requirement in this type of course),
attendance and active discussion at presentations of other students,
pass the exam, i.e., prove the knowledge of basic concepts presented during the course.

Previous runs

XP36VPD – Selected parts of data mining, a very similar content (despite its title it covered ML topics too), running since 2015.

Table of Contents