Teemu Roos

Ph.D., Assistant Professor

Leader of the Information, Complexity and Learning research group

Department of Computer Science and
Helsinki Institute for Information Technology HIIT
Exactum building, A322
PO Box 68
FI-00014 University of Helsinki
Finland


contact: teemu.roos at cs.helsinki.fi

What's up?

NEW: Our fast approximate nearest neighbor algorithm is available as an optimized C++ package with Python bindings. To my knowledge, this is the fastest available algorithm for high accuracy queries (see our forthcoming IEEE Big Data paper below).
C++ library (with Python bindings) | benchmarks

As of 1/2017, I'll be appointed as an Associate Professor (tenure-track) at the Department of Computer Science, University of Helsinki.

The 9th Workshop on Information Theoretic Methods in Science and Engineering (WITMSE) will be held in Helsinki on 19–21 September 2016. web page.

My traveling/conference schedule in 2016:

Since November 2013, following Prof Juho Rousu, I am the host of the Helsinki Distinguished Lecture Series on Future Information Technology.

I have been appointed from 4/2013 onwards as an Assistant Professor (tenure-track) at the Department of Computer Science, University of Helsinki.

Recent conference and workshop involvement (program committee or equivalent): AISTATS-2011/2014/2015, ECML/PKDD-2009/2012, ICDM-2014, IJCAI-2013/2015, NIPS-2014/2015/2016, PGM-2008/2012/2014, UAI-2008/2009/2010/2011/2012/2013/2014/2015/2016.

Teaching

In the academic year 2016–2017, I will teach the following courses:

I am the instructor of the Undergraduate Research Track (tutkijalinja).

I currently supervise:

  • Joonas Miettinen (PhD student; 1st year) co-supervised with Dr Janne Pitkäniemi (Finnish Cancer Registry)
  • Beenish Qaiser (PhD student; 1st year) co-supervised with Dr Anu Loukola (FIMM)
  • Ville Hyvönen (PhD student, 2nd year)
  • Janne Leppä-aho (PhD student, 2nd year)
  • Pedram Daee (PhD student, 3rd year) co-supervised with Prof Sami Kaski
  • Yuan Zou (PhD student, 6th year)
  • Teemu Pitkänen (MSc student)
  • Elaine Zosa (MSc student)
  • Joe Niemi (MSc student)
  • Tommi Jalkanen (MSc student) co-supervised with Dr Jutta Jokiranta (Theology)
  • Olli Orajärvi (MSc student)

Past students:

  • Jussi Määttä (PhD 2016)
  • Antti Takalahti (MSc 2016; co-supervised with Prof Edward Haeggström (Physics)
  • Risto Tuomainen (MSc 2016)
  • Simo Linkola (MSc 2016)
  • Kaj Sotala (MSc 2015)
  • Quan Nguyen (MSc 2015)
  • Peter Hedman (MSc 2015), co-supervised with Prof Jaakko Lehtinen (Aalto/NVIDIA Corp) Acad. Assoc. for Maths and Natural Sciences (MAL) MSc Thesis Award
  • Arttu Modig (MSc 2014), co-supervised with Prof Jouko Lampinen (Aalto)
  • Janne Leppä-aho (MSc 2014), co-supervised with Prof Jukka Corander
  • Henning Lübbers (MSc 2012), co-supervised with Prof Jyrki Kivinen
  • Anupam Arohi (MSc 2011)
  • Yuan Zou (MSc 2011)
  • Teemu Pulkkinen (MSc 2011)
  • Toni Merivuori (MSc 2009)
  • Lari Latvala (MSc 2009), co-supervised with Prof Jouko Laasasenaho

Students interested in MSc/PhD thesis topics related to information theory, statistical modeling, machine learning, artificial intelligence, and digital humanities are welcome to contact me by e-mail (but see disclaimer below).

Disclaimer: If you currently not enrolled at the University of Helsinki, please do not send me e-mail, but contact the Department of Computer Science for information about the application process. Unfortunately I cannot reply to all e-mail inquiries.

Research

``Your act was unwise,'' I exclaimed ``as you see by the outcome.''
He solemnly eyed me. ``When choosing the course of my action,''
said he, ``I had not the outcome to guide me.''
[Ambrose Bierce]

I am the leader of the Information, Complexity and Learning (ICL) research group, which is a part of the CoSCo research group led by Prof Petri Myllymäki. I'm also affiliated with the Academy of Finland funded Centre of Excellence COIN.

Topics of my interest include the theory and applications of

  • machine learning and big data
  • computational statistics
  • probabilistic graphical models
  • information theory
  • digital humanities

Recent and ongoing work:

  1. V. Hyvönen, T. Pitkänen, S. Tasoulis, E. Jääsaari, R. Tuomainen, L. Wang, J. Corander, and T. Roos. Fast nearest neighbor search through sparse random projections and voting, to appear in 2016 IEEE International Conference on Big Data (IEEE Big-Data 2016), Washington DC, Dec. 5–8. C++ library (with Python bindings) | benchmarks

  2. J. Leppä-aho, J. Pensar, T. Roos, and J. Corander (submitted). Learning Gaussian graphical models with fractional marginal pseudo-likelihood, in revision, arXiv:1602.07863

  3. Y. Zou and T. Roos (2017). On model selection, Bayesian networks, and the Fisher information integral, to appear in New Generation Computing, 35(1) (Special Issue on AMBN 2015), January 2017.

  4. Y. Zou and T. Roos (2016). Sparse logistic regression with logical features, in J. Bailey, L. Khan, T. Washio, G. Dobbie, J. Z. Huang, and R. Wang (editors), Proc. 20th Pacific Asia Conference on Knowledge Discovery and Data Mining (PAKDD-2016), Lecture Notes in Artificial Intelligence 9652, Springer, pp. 316–327.

  5. T. Heikkilä and T. Roos, (2016). Thematic Section on Studia Stemmatologica, Digital Scholarship in the Humanities 31(3):520–522, doi:10.1093/llc/fqw038.

  6. T. Roos (2016). Minimum Description Length Principle, in Sammut, C. and Webb G.I. (eds), Encyclopedia of Machine Learning and Data Mining, 2016.

  7. L. Wang, S. Tasoulis, T. Roos, and J. Kangasharju (2016). Kvasir: Scalable provision of semantically relevant web content on big data framework, to appear in IEEE Transactions on Big Data.

  8. Y. Zhao, S. Tasoulis, and T. Roos (2016). Manifold visualization via short walks, in E. Bertini, N. Elmqvist, and T. Wishchgoll (editors), Eurographics Conference on Visualization (EuroVis-2016), The Eurographics Association, pp. 85–89, DOI:10.2312/eurovisshort.20161166

  9. J. Määttä and T. Roos (2016). Maximum parsimony and the skewness test: A simulation study of the limits of applicability, PLOS ONE 11(4):e0152656. C++/Python code

  10. J. Määttä, D.F. Schmidt, and T. Roos (2016). Subset selection in linear regression using sequentially normalized least squares: Asymptotic theory, Scandinavian Journal of Statistics 43(2):382–395.

  11. J. Määttä and T. Roos (2016). Robust sequential prediction in linear regression with Student's t-distribution, in Proc. 14th International Symposium on Artificial Intelligence and Mathematics (ISAIM-2016).

  12. J. Tehrani, Q. Nguyen, and T. Roos, (2016). Oral fairy tale or literary fake? Investigating the origins of Little Red Riding Hood using phylogenetic network analysis, Digital Scholarship in the Humanities 31(3):611–636.

Selected publications (full list, Google Scholar, DBLP):

  1. K. Watanabe and T. Roos, (2015). Achievability of asymptotic minimax regret by horizon-dependent and horizon-independent strategies, Journal of Machine Learning Research 16(Nov):2357–2375.

  2. Q. Nguyen and T. Roos, (2015). Likelihood-based inference of phylogenetic networks from sequence data by PhyloDAG, in Proc. 2nd International Conference on Algorithms for Computational Biology (AlCoB-2015), LNBI 9199, Springer, pp. 126–140. R/C++ code

  3. A. Barron, T. Roos, and K. Watanabe, (2014). Bayesian properties of normalized maximum likelihood and its fast computation, in Proc. IEEE International Symposium on Information Theory (ISIT-2014), IEEE Press, pp. 1667–1671.

  4. M. Sherman, G. Clark, Y. Yang, S. Sugrim, A. Modig, J. Lindqvist, A. Oulasvirta, and T. Roos, (2014). User-generated free-form gestures for authentication: security and memorability, in Proc. 12th International Conference on Mobile Systems, Applications, and Services (MobiSys-2014), ACM Press, pp. 176–189.

  5. A. Carvalho, T. Roos, A. Oliveira, and P. Myllymäki, (2011). Discriminative learning of Bayesian networks via factorized conditional log-likelihood, JMLR 12(Jul):2181–2210.

  6. J. Rissanen, T. Roos, and P. Myllymäki, (2010). Model selection by sequentially normalized least squares, Journal of Multivariate Analysis 101:4, 839–849.   preprint | R code

  7. T. Roos and T. Heikkilä, (2009). Evaluating methods for computer-assisted stemmatology using artificial benchmark data sets, Literary and Linguistic Computing, 24:4, 417–433, doi:10.1093/llc/fqp002. abstract | data-sets

Past Events

I have held visiting positions at CWI Amsterdam (2003 & 2005), UC Berkeley (2008, 2015), MIT (2010), University of Cambridge (2012), and the Finnish Institute in Rome (2013).

Between September 21–November 30, 2015, I was a Visiting Scholar at UC Berkeley, USC and UCSD.

Workshop on Advances in High Dimensional Big Data in association with the IEEE Big Data 2015 conference; co-chairs Tasoulis, Roos & Corander. call for papers (due August 30)

The 8th Workshop on Information Theoretic Methods in Science and Engineering (WITMSE) was held in Copenhagen on 24–26 June 2015; co-chairs Harremoës, Forchhammer, Roos & Myllymäki. web page.

My traveling/conference schedule in 2015:

I was a member of the organizing committee of the 11th Conference of the European Society for Textual Scholarship (ESTS-2014). The conference was held in Helsinki on Oct 30-Nov 1, 2014.

We organized the 7th Workshop on Information Theoretic Methods in Science and Engineering (WITMSE) in Hawaii on July 5-8, 2014, right after the ISIT symposium. website

In October–December 2013, I was a Fellow at the Finnish Institute in Rome (Villa Lante).

Associated with our CHI-2013 paper, we are lauching a web server for computing the information capacity from your own motion capture data: infocapacity.hiit.fi. Please ask for a beta testing account.

My traveling/conference schedule in 2014:

I was a co-chair of WITMSE-2013 in Tokyo, Japan, August 26–29.

My traveling/conference schedule in 2013:

Our paper "Information capacity of full-body movements" gets a Best Paper Honorable Mention Award at CHI-2013. If you are planning to attend the conference, please visit our fun interactive Kinect-based demo!

Special issue on selected papers from PGM-2010 in the International Journal of Approximate Reasoning (editors Roos, Myllymäki, Jaakkola): link to editorial and articles.

I served as a member of the Senior Program Committee of UAI-2012 and an Area Chair at ECML-PKDD 2012.

In January–April 2012, I was a Visiting Fellow at the University of Cambridge.

I was an external evaluator at the PhD thesis defense of Thomas Toftkjær at Aarhus University on January 10, 2012.

My traveling/conference schedule in 2012:

The Academy of Finland Centre of Excellence COIN starts in 1/2012.

We organized the 4th Workshop on Information Theoretic Methods in Science and Engineering in Helsinki, on August 7–10, 2011, right after ISIT 2011. web pages

The 5th Brazilian Conference on Statistical Modelling in Insurance and Finance was held in Maresias, Brazil, on April 10–15, 2011. I gave a two-day short course on MDL. lecture notes | slides (day1) | slides (day2)

Tuomas Heikkilä, Petri Myllymäki and I organize a series of stemmatology workshops in Helsinki and elsewhere in 2010–2012. web pages

Cambridge, UK, March 2011.
My traveling/conference schedule in 2011:

4/2011: I was conferred the title Adjunct Professor (in Finnish, dosentti) by the Faculty of Science, and appointed as a senior reseacher at HIIT.

I was invited to the senior program committee of UAI-2011.

The Academy of Finland has graciously decided to fund me under a postdoctoral researcher's project.

Petri Myllymäki, Tommi Jaakkola, and I were the program committee co-chairs of the 5th European Workshop on Probabilistic Graphical Models (PGM-2010) in Helsinki, September 13–15, 2010. web pages

In February–April 2010, I visited Prof. Tommi Jaakkola's group at MIT, Boston.

My traveling schedule in 2010:

I got the ERCIM (European Research Consortium for Informatics and Mathematics) 2009 Cor Baayen Award.

Pisa, November 2009.
My traveling schedule in 2009 (a lot of ITs!):

In Fall 2009, I lectured the new course Information-Theoretic Modeling (4 cr) and Information-Theoretic Modeling Project (2 cr).

The University of Helsinki has granted funding to project STAM (Algorithmic Methods in Stemmatology) for the years 2009–2011. project website | "Computer programs can do wonders"

The Finnish Cultural Foundation has awarded a Science Workshop grant (EUR 200,000) on stemmatology for the years 2009–2010. announcement (w/ fanfares) (in Finnish)

As of August 2008, I have been appointed as post-doctoral researcher at HIIT for three years.

Berkeley, April 2008.
In January–April 2008, I visited UC Berkeley (Prof. Bin Yu's group) and ICSI.

During the Fall term 2007 I lectured the Three Concepts: Information course.

I defended my Ph.D. thesis "Statistical and Information-Theoretic Methods for Data Analysis" on June 9, 2007. The opponent was Prof. Alon Orlitsky (UCSD). Pre-examiners were Prof. Ioan Tabus (Tampere UT) and Prof. Tommi Jaakkola (MIT). electronic version (summary part).

I received a Ph.D. degree (in Computer Science) from the University of Helsinki in 2007. I was supported by HeCSE (Helsinki Graduate School in Computer Science and Engineering). My supervisors were Prof. Henry Tirri (on industrial leave), and Prof. Petri Myllymäki. In addition to computer science I have minors in mathematics and philosophy (see a list of finished courses).

Earlier, I have done some work on mobile device positioning. For scientific publications, see list of publications. For working products, go to Ekahau.

Reason

I am married to the loveliest girl in the world, the light of my life. ''You are the reason I am. You are all my reasons.''

Since July 31st 2003, the universe revolves around a boy. Since March 1st 2007, we have two boys!

Other

Take a look at (old) Cosco papers visualized using the Similarity Metric of Vitányi and Cilibrasi and multidimensional scaling (Sammon mapping).

Play a game of Rock, Paper, Scissors: rock beats scissors, paper beats rock, and scissors beat paper. Even such a simple game offers some theoretically interesting problems: Can one predict the other player's choice? What is the best strategy against a good opponent? (Yes, the computer could cheat, but I promise it doesn't.) Such questions were considered by Claude Shannon in the 1950s; see a modern variation of his 'Mind-Reading Machine' based on data-compression (CTW).

Choose one: Your choice: My choice: You Me Draw

Last updated on November 18, 2016