Seminar on Intelligent Systems

58309104
3
Algorithms and machine learning
Advanced studies
Year Semester Date Period Language In charge
2011 spring 17.01-25.04. 3-4 English Petri Myllymäki

Lectures

Time Room Lecturer Date
Mon 16-18 C220 Petri Myllymäki 17.01.2011-21.02.2011
Mon 16-18 C220 Petri Myllymäki 14.03.2011-25.04.2011

Information for international students

If there are any non-Finnish speaking students, the seminar will be held in English, otherwise the presenter can decide between English and Finnish.

General

The topic of the seminar is  Challenges in Intelligent Systems. Recently, there have been numerous public challenges or competitions related to Intelligent Systems, and in this seminar we will explore some of them in more detail. For each challenge, the goal is to understand

  • What is this challenge about? Why is solving it important?
  • What are the main approaches used for solving the challenge?

There is no overlap with the previous seminars on Intelligent Systems, so you can take this seminar even if you attended one of the previous ones.

Completing the course

To earn the available credit points, you need to

  1. Browse the Internet for a suitable Challenge. In principle any challenge related to some area of Intelligent Systems or Artificial Intelligence is OK (most of the Challenges address Machine Learning or Data Mining problems), as long as there is sufficient public documentation available for the challenge itself, and the various approaches used for solving it. When you have have found a suitable challenge, send a suggestion to the seminar instructor. If there are two students suggesting the same challenge, or two challenges that are very similar to each other, only the first suggestion is accepted.
  2. Give two talks at the seminar. The first talk (in Period III) is for presenting the Challenge. In the second talk (in Period IV) you need to describe the solutions used: give both a general overview of the solutions, and then focus in more detail on at least  three different approaches for solving the Challenge (if the three best solutions are using different approaches, you can focus on them, but if e.g. the first two are very similar, you need to take another solution for closer inspection).
  3. Attend the talks of the other students and participate in the discussion (one missed event is acceptable).
  4. Write a seminar report containing a paragraph (of about 10 lines at least) on each of the talks you attended, summarizing the main points you learned from the talk.

You do not have to write a seminar paper on your Challenge, but you need to deliver copies of your slides after each of the talks, and the quality of the slides is one aspect in the grading of your performance. All in all, your grade will depend upon

  1. Quality of talk I, how  well did you manage to describe the Challenge and the necessary background.
  2. Quality of talk II, how well  did you manage to describe the solutions: a poor presentation only describes the minimum of three different approaches on a superficial level, while a good presentation incudes all the suggested solutions, organized into a logical hierarchy with similar approaches grouped together, and presents the methods used in an understandable manner, demonstrating good understanding of the methodologies used.
  3. Quality of your seminar report.

 The first talks should be 20 minutes long, and we will have a short break for discussion between the talks. There is a separate session for a more general discussion on March 14 (After Kasari's talk), where the idea is to discuss together about what we learned so far about about giving a scientific presentation: base on the first talks, what makes a good (bad) talk? Please listen carefully, and use the feedback for improving your second talk.

In the second talks, which are 25 minutes long, you should not continue directly where you left off at the end of your first talk, but try to make the second talk also more or less indpendent: the introduction of the challenge should of course be relatively short, but try to exploit the feedback received and focus on the main points of the problem, especially considering the methodological issues that are the main point of the second talk.

 

Seminar Report

You should include one entry in the final report for each of the talks you heard (there are 13 attendees each giving two talks, so each of you should have 2 x 12 entries in your seminar report). Each entry should contain three elements:

  1. A short (a few sentences/one paragraph) summary of the contents of the talk: What was it about? What did you find interesting? What did you learn?
  2. A short evalluation of the technical merits of the presentation: did you get the impression that the speaker had acquired the necessary information, was the technical contents deep/wide enough, was the speaker capable of answering questions properly? In addition to your verbal evaluation, give a grade on a scale (+,++,+++), where "++" denotes an average performance in the seminar, "+" below average and "+++" better than average.
  3. A short evaluation of the quality of the presentation: was the talk well structured, how was the quality of the verbal presentation, or the quality of the supplementary material (slides, videos, etc.)? Did you understand what the speaker was trying to say? In addition to your verbal evaluation, give a grade on a scale (+,++,+++), where "++" denotes an average performance in the seminar, "+" below average and "+++" better than average.

You can either structure your report  by discussing the first talks first (in chronological or some other order), and then the second talks, or you can group the first and second talk of the same presenter together, whatever feels more natural for you.

 

Participants and Topics

 

Schedule (tentative)

Monday 17.01.: First meeting, organization of the seminar.

Monday 24.01.: Q&A, fixing the schedule (not a compulsory session if the topic has already been selected.).

Monday 31.01.: Q&A about the talks (not a compulsory session)

Monday 07.02.: First talks of Suvi Hiltunen, Niklas Jahnsson, Guowei Lv and Jussi Karppinen

Monday 14.02.: First talks of Kalervo Oikarinen, Tommi Rönkönharju, Ziran Wang and Lilu Xu

Monday 21.02.: First talks of Amin Alizadeh, Mika Wahlroos, Tao Xu, Wenqing Zhou and Oskar Gross

Monday 28.02.: No seminar (period break)

Monday 07.03.: No seminar (period break)

Monday 14.03.: First talk of Melissa Kasari. General discussion on all the first talks: what did we learn about giving presentations, what type of elements make a complex topic more easy to understand after a relatively short presentation?

Monday 21.03.: No seminar

Monday 28.03.: Second talks of Niklas Jahnsson, Jussi Karppinen,  Kalervo Oikarinen and Tommi Rönkönharju.

Monday 04.04.: Second talks of Ziran Wang, Lilu Xu and Amin Alizadeh.

Monday 11.04.: Second talks of Mika Wahlroos, Tao Xu and Wenqing Zhou

Monday 18.04.: Second talks of Suvi Hiltunen,  Guowei Lv, Melissa Kasari and Oskar Gross .

Monday 25.04.: No seminar (Easter)

Friday 29.04.: Deadline for the seminar report at 23:59 (midnight), delivery by email to the seminar instructor.

 

Literature and material

Finding a suitable Challenge is part of the seminar work, but the task is not too difficult - try e.g. googling "data analysis" or "machine learning" with "challenge" or "competition" - but here are some easy starting points for potential candidates: