Hauptseminar Computational Aspects of Machine Learning - Winter 16
- Term
- Winter 16
- Lecturer
- Kilian Röhner, Moritz August, Paul Cristian Sarbu
- Time and Place
- Wed. 10am till 12pm room MI 02.07.023 starting from 26. October (see schedule below)
- Audience
- Students from Bachelor {Informatics, Information Systems, Games Engineering} (IN0014), Master {Informatics, Games Engineering} (IN2107) and Master Computational Science and Engineering (IN2183)
- Tutorials
- -
- Exam
- -
- Semesterwochenstunden / ECTS Credits
- 2 SWS / 4 Credits
- TUMonline
- https://campus.tum.de/tumonline/LV.edit?clvnr=950267201
Contents
News
- Preliminary Meeting: 28th of June, 12pm, MI 02.07.23
- Take a look at the Coursera Machine Learning course. If you are new to the field, the online course is a good way to get an overview.
Description
Machine Learning is a rapidly growing area of research on the intersection of applied mathematics, informatics, and computational science. With advances in the machine learning theory and algorithms as well es with increasing amount of data and complexity of the models, development of fast, efficient and scalable algorithms increasingly gains importance. Hereby the range of applied techniques spreads from exploiting embarrassingly parallel tasks in a data-centric fashion to approximation methods with satisfying error limits.
In the seminar we are going to focus on the advanced methods for machine learning with particular interest in handling large-scale problems. While some topics would deal with the complete learning algorithms, others would focus on efficient solution of subtasks common for many different algorithms, e.g. nearest neighbors search or MCMC sampling.
Schedule
Preliminary schedule of the presentation
| Date | Forename | Surname | Topic | Advisor | Chair | Reviewer 1 | Reviewer 2 | |
|---|---|---|---|---|---|---|---|---|
| 26.10.15 | Julian | Landesberger | Large Scale Bayesian Inference | Moritz | Daniel L. | Yue C. | James B. | |
| 02.11.15 | Daniel | Lehmberg | Online Learning | Kilian | Jait D. | Jait D. | Ingo M. | |
| 16.11.15 | Andrei | Costinescu | Temporal Difference Learning | Paul | Julian L. | Oleksandr M. | Yehor Y. | Julian L. |
| 23.11.15 | Yehor | Yudin | Temporal Data Mining | Paul | Andrei C. | Andrei C. | Oleksandr M. | |
| 30.11.15 | Ingo | Mayer | Multi-Agent Systems and their Learning Process | Paul | Oleksandr M. | Yehor Y. | Jait D. | |
| 07.12.15 | James | Browne | Large Scale Graph Mining | Kilian | Ingo M. | Jait D. | Daniel L. | |
| 14.12.15 | Jait | Dixit | Hardware for Deep Learning | Moritz | Yehor Y. | Oleksandr M. | Julian L. | James B. |
| 11.01.16 | Oleksandr | Melkonyan | Speeding Up Deep Learning Algorithms | Moritz | James B. | Ingo M. | Andrei C. | Daniel L. |
We will cover the following topics in the seminar, also participants may suggest their own topics if those suit the goals of the seminar.
- Parallelisation paradigms and parallel performance models
- Large scale Bayesian inference
- Large scale kernel learning
- Speeding-up deep learning models
- Speeding-up deep learning training
- Resource-aware/constrained machine learning methods
- Online learning
- Real-time data mining with guaranteed throughput
- Data mining with sparse grids
- Temporal data mining
- Large scale graph mining
- Machine learning as a black box solution
- Cloud-based machine learning
- Temporal difference learning
- Manifold learning
- ...
Materials
Prerequisites
The seminar is intended for computer science bachelor students in advanced semesters and the master students with focus on computer science, computational science and engineering, or robotics and artificial intelligence.
If you want to participate, you are expected to have knowledge of fundamentals of applied mathematics, such as linear algebra, probability theory, calculus, and convex optimisation. As we will cover advanced topics, some general understanding of the basic concepts of machine learning is required.
Moreover, you should have programming experience in at least one programming language. Understanding of basic concepts of computational science is desirable, so if you know what SIMD, MISD, NUMA, MPI, GPGPU, and HDF5 mean, you are on the safe side.
Application
If you feel like a data science unicorn we are looking for, please follow the link to application (is available until July 6th, same deadline as the matching system) and send us the information about your background and experience, 3 preferred topic (ranked) or your own topic suggestions, as well as your expectations from the seminar. Please use this application form.
Additionally, you are required to register for the seminar in the matching system.
Resources
- IEEE Guide for Reviewers if you don't know how to review other student's papers
- Mathematical Writing the script of the timeless Stanford lecture by Donald Knuth (also available on youtube)
Organisational Details
- Weekly sessions. Day, time, and room tba.
- Talk: 45 minutes + 15 minutes discussion
- Extended abstract: 1 page article document class with motivation, key concepts and results
- Paper: min. 5 pages max. 6 pages in IEEE format
- Language: English
- max. 14 Participants
- Blind peer-review process
- Session chairing
Dates
Dates to keep in mind:
- July 1st till 6th: Application and registration in the matching system of the department
- after July 23th: Notification of participants, topic discussions, etc.
- Week starting with October 17th: first session, first talk.
Deadlines
- 1 week before the talk: submission of an extended abstract
- The days of 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
LaTeX
- Please use the IEEE Template.
