2016

Prof. Debotosh Bhattacharjee

Jadavpur University, Kolkata, India

13.6. – 13.7.2016
Research visit.

Prof. Mirjana Ivanović, assoc. prof. Vladimir Kurbalija, assist. prof. Miloš Savić

University of Novi Sad, Sebia

12.6. – 17. 6. 2016
Research collaboration on bilateral project "Intelligent computer techniques for improving medical detection, analysis and explanation of human cognition and behavior disorders."

2015

Prof. Dr. Snežana Šćepanović

University Mediteran, Monte Negro

8.12. – 11.12.2015
Research collaboration on bilateral project "Development of new e-learning models for game-based  learning using mobile technologies".

Prof. dr. Zikrija Avdagić, Samir Omanović

University of Sarajevo, BIH

15.11 - 18.11.2015
Research collaboration on Computer based modeling in bioinformatics for gene based cancer classification focused on reliability and machine learning.

Doc. dr. Josip Musić, Dr. Ivo Stančić

University of Split, Croatia

4.10 - 7.10.2015
Research collaboration on Supervised and unsupervised learning from imbalanced datasets for assistance in movement of persons with low vision.

Prof. Dr. Ujjwal Maulik

University Kolkata, India

28.9. - 1.10.2015
Research collaboration on machine learning and data mining.

Petar Radunović, Tijana Vujičić, Nađa Žarić

University Mediteran, Monte Negro

21.9. – 24.9.2015
Research collaboration on bilateral project "Development of new e-learning models for game-based learning using mobile technologies".

Prof. Dr. Mirjana Ivanović

University of Novi Sad, Serbia

8.6 - 13.6.2015
Research collaboration on education and text mining technolgies.

2014

dr. Josip Musić, dr. Ivo Stančić and mag. Ante Panjkota

University of Split, Croatia

9.12. - 13.12.2014
research collaboration on Supervised and unsupervised learning from imbalanced datasets for assistance in movement of persons with low vision.

Tripo Matijević, Tijana Vujičić and Nađa Žarić

Univerzitet Mediteran, Monte Negro

15.12. – 18.12.2014
Research collaboration on bilateral project "Development of new e-learning models for game-based learning using mobile technologies".

2013

prof. dr. Zikrija Avdagić, dr. Aida Saračević and mag. Dino Kečo

University of Sarajevo, Serbia

27.5. − 29.5.2013
computer aided lung cancer classification of mutated EGFR exons using artificial intelligence methods.

2011

prof. dr. Petr Savicky

University of Prague, Czech Republic

30.5. − 10.6.2011
research collaboration on Machine Learning of Imbalanced Data.

dr. Pedro Pereira Rodrigues

University of Porto, Portugal

29.6. − 8.7. 2011
research collaboration on electricity load forecasting supported by prediction explanation and prediction reliability estimates.

prof. dr.  João Gama

University of Porto, Portugal

4.7 − 9.7. 2011
research collaboration on electricity load forecasting supported by prediction explanation and prediction reliability estimates.

prof. dr.  Nenad Filipović, Miloš Radović and dr. Aleksandar Peulić

University of Kragujevac, Serbia

20.4 − 22.4.2011
research collaboration on use of machine learning for modeling coronary artery disease.

prof. dr.  Tatjana Zrimec

University of New South Wales, Australia

3.3. − 20.7.2011
research collaboration on machine learning from medical image data.

Ercan Canhasi

University of Priština, Kosovo

15.1. − 30.6.2011
Multidocument summarization based on multilayered graphs.

2010

prof. dr. Petr Savicky

University of Prague, Czech Republic

23.8.2010 − 3.9.2010,
11.11.2010 − 23.11.2010,
research collaboration on Learning in Imbalanced Data.

dr. Pedro Pereira Rodrigues and Raquel Sebastião (PhD student)

University of Porto, Portugal

13.9.2010 − 22.9.2010,
research collaboration on data streams mining and electricity load forecasting.

Ercan Canhas, MSc

University of Priština, Kosovo

1.1.2010 − 15.9.2010,
PhD scholarship funded by EU, research in text mining.

2009

Petr Savicky

10. 12. 2009 - 18.12.2009
11. 9. 2009 - 17. 9. 2009

2008

Petr Savicky

19. 11. 2008 - 26. 11. 2008


Title: Arithmetic circuits of Bayesian networks with noisy-or dependencies
presentation by J. Vomlel and P. Savicky
Bayesian networks (BN) are quite a general type of models for representing distributions on several variables, among which we have only a limited number of dependencies. We consider only BN on discrete variables. Such BN may be represented by a multilinear polynomial. If this polynomial may be evaluated using a small arithmetic circuit, it may be used as an alternative way of efficient inference in the distribution described by the BN. The talk will describe the polynomial, briefly discusses some of the known methods for constructing circuits for this polynomial and then we present an experimental comparison between two ways of preprocessing networks containing noisy-or dependencies (parent divorcing and tensor rank one decomposition) before constructing the circuit. Good properties of the latter are supported both theoretically and experimentaly.

2007

Pedro Pereira Rodrigues

5.10.2007 – 13.10.2007

Cláudia Camila Rodrigues Pereira Dias

5.10.2007 – 13.10.2007

Raul Fidalgo

september 2007 – december 2007

Harry Wechsler

22.6.2007 – 26.6.2007

2006

prof. dr. Joao Gama

researcher at LIACC, the Laboratory of Artificial Intelligence and Computer Science of the University of Porto.

6.11.2006 - 11.11.2006
Title: Learning Forest of Trees from Data Streams
Abstract: This lecture presents a system for induction of forest of functional trees from data streams able to detect concept drift.  The Ultra Fast Forest of Trees (UFFT) is an incremental algorithm, that works online, processing each example in constant time, and performing a single scan over the training examples. It uses analytical techniques to choose the splitting criteria, and the information gain to estimate the merit of each possible splitting-test.  For multi-class problems the algorithm grows a binary tree for each possible pair of classes, leading to a forest of trees.  Decision nodes and leaves contain naive-Bayes classifiers playing different roles during the induction process.  Naive-Bayes in leaves are used to classify test examples, naive-Bayes in inner nodes can be used as multivariate splitting-tests if chosen by the splitting criteria, and used to detect drift in the distribution of the examples that traverse the node. The use of naive-Bayes classifiers at leaves to classify test examples, the use of splitting-tests based on the outcome of naive-Bayes, and the use of naive-Bayes classifiers at decision nodes to detect drift are directly obtained from the sufficient statistics required to compute the splitting criteria, without no additional computations. This aspect is a main advantage in the context of high-speed data streams.  This methodology was tested with artificial and real-world data sets. The experimental results show a very good performance in comparison to a batch decision tree learner, and high capacity to detect and  react to drift.

Rita Ribeiro

researcher at LIACC, the Laboratory of Artificial Intelligence and Computer Science of the University of Porto.

6.11.2006 - 11.11.2006
Title: R-PREV: A Rule-based framework for the Prediction of Rare Extreme Values
Abstract: In this presentation it will be described a rule learning method that obtains models biased towards a particular class of regression tasks. These tasks have as main distinguishing feature the fact that the main goal is to be accurate at predicting rare extreme values of the continuous target variable. Many real-world applications from scientific areas like ecology, meteorology, finance,etc., share this objective. Most existing approaches to regression problems search for the  model parameters that optimize a given average error estimator (e.g. mean squared error). This means that they are biased towards achieving a good performance on the most common cases. The motivation for our work is the claim that being accurate at a small set of rare cases requires different error metrics. Moreover, given the nature and  relevance of this type of applications an interpretable model is usually of key importance to domain experts, as predicting these rare events is normally associated with costly decisions. Our proposed  system R-PREV obtains a set of interpretable regression rules derived from a set of bagged regression trees using evaluation metrics that bias the resulting models to predict accurately rare extreme values. We provide an experimental evaluation of our method confirming the advantages of our proposal in terms of accuracy in predicting rare extreme values.