Machine learning: user modeling, implicit feedback like eye movements, sparse models, decision trees, dicretization of continuous variables, online learning, matrix factorization models, and topic models. Technologies that process and store large-scale data (map-reduce), data structures, and I/O -models of computational cost.
There is a bibtex file of the following publications: publications.bib.
Publications in chronological order. I've marked papers with more interesting content with ★.
Jussi Kujala, Tapio Elomaa: On Following the Perturbed Leader in the Bandit Setting. Algorithmic Learning Theory (ALT) 2005: LNCS volume 3734 p. 371-385.
Tapio Elomaa, Jussi Kujala, Juho Rousu: Approximation Algorithms for Minimizing Empirical Error by Axis-Parallel Hyperplanes. European Conference on Machine Learning (ECML) 2005: LNCS volume 3720 p. 547-555.
Tapio Elomaa, Jussi Kujala, Juho Rousu: Practical Approximation of Optimal Multivariate Discretization. International Symposium on Methodologies for Intelligent Systems (ISMIS) 2006: LNCS volume 4203 p. 612-621.
Jussi Kujala, Tapio Elomaa: Poketree: A Dynamically Competitive Data Structure with Good Worst-Case Performance. International Symposium on Algorithms and Computation (ISAAC) 2006: LNCS volume 4288 p. 277-288.
Jussi Kujala, Tapio Elomaa: Improved Algorithms for Univariate Discretization of Continuous Features. Principles and Practice of Knowledge Discovery in Databases (PKDD) 2007: LNCS volume 4702 p. 188-199.
Jussi Kujala, Tapio Elomaa: Following the Perturbed Leader to Gamble at Multi-Armed Bandits. Algorithmic Learning Theory (ALT) 2007: LNCS volume 4754 p. 158-172.
Timo Aho, Tapio Elomaa, Jussi Kujala: Reducing Splaying by Taking Advantage of Working Sets.
Workshop on Experimental Algorithms (WEA) 2008: LNCS volume 5038 p. 1-13.
Timo Aho, Tapio Elomaa, Jussi Kujala: Unsupervised Classifier Selection Based on Two-Sample Test. Discovery Science (DS) 2008: LNCS volume 5255 p. 28-39.
Jussi Kujala, Tapio Elomaa: The Cost of Offline Binary Search Tree Algorithms and the Complexity of the Request Sequence. Theoretical Computer Science: volume 393 issue 1-3 p. 231-239, 2008.
Jussi Kujala, Tapio Elomaa: Ranking the Uniformity of Interval Pairs. ECML-PKDD 2008: LNAI volume 5211 p. 640-655.
★ Jussi Kujala: Assembling Approximately Optimal Binary Search Trees Efficiently Using Arithmetics. Information Processing Letters (IPL): volume 109 issue 16 p. 962-966, 2009. There is also a short implementation of this algorithm in C: awobst.tar.bz2.
★ Jussi Kujala, Timo Aho, Tapio Elomaa: A Walk from 2-norm SVM to 1-norm SVM. In proceedings of IEEE ICDM'09 p. 836-841. You can find a patch to liblinear 1.33 here.
Tapio Elomaa, Jussi Kujala: Covering Analysis of the Greedy Algorithm for Partial Cover. Algorithms and Applications 2010 p. 102-113.
Peter Auer, Zakria Hussain, Samuel Kaski, Arto Klami, Jussi Kujala, Jorma Laaksonen, Alex P. Leung, Kitsuchart Pasupa, John Shawe-Taylor: Pinview: Implicit Feedback in Content-Based Image Retrieval. In JMLR Workshop and Conference Proceedings (Workshop on Applications of Pattern Analysis) 2010.
★ Samuel Kaski, Arto Klami, Jussi Kujala (editor), Shiau Hong Lim, Chiwei Wang. Report on potential improvements obtainable by data fusion. PinView FP7-216529 Deliverable D2.1, 2010. Despite the name contains material on e.g. performing matrix factorization with stochastic gradient and on how much data is needed to achieve a desired accuracy.
This manuscript studies a topic model which is not probabilistic but optimizes an objective function that promotes sparseness (Dirichlet prior in LDA is linearised near zero). The paper is here. There's also reference code here. Note that software such as GibbsLDA++ is good on single core, and more likely stable ;) Also, PLSI works well if the number of parameters of the model is not large compared to data size, and it parallelizes well.
Jussi Kujala: Aktiojärjestelmien toteuttaminen EJB-ympäristössä (in English: Implementation of action systems in EJB environment). This is my M.Sc.(Eng.) Thesis.
Jussi Kujala: Improvements to Online Learning Algorithms with Applications to Binary Search Trees. This is the introduction to my Ph.D. Thesis which also included five articles in addition to this introduction.