Professor Wojciech Penczek elected a Corresponding Member of the Polish Academy of Sciences

ZOPAN WPenczek

Professor Wojciech Penczek, Director of ICS PAS has been elected a Corresponding Member of the Polish Academy of Sciences. The 140 sesion of General Assembly of the Polish Academy of Sciences took plase on December 5. In accordance with the Regulations on the selection of Academy members, candidates for members should be distinguished by their special scientific achievements, environmental authority and good repute.



Picture 1: Private Archive
Picture 2: Jarosław Deluga-Góra / PAS

Work of Piotr Przybyła was accepted at AAAI 2020


Work of Piotr Przybyła on automatic methods that can detect online documents of low credibility was accepted at AAAI 2020.

AAAI conference series is to promote research in artificial intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers in AI and its affiliated disciplines. AAAI-20 is the Thirty-Fourth AAAI Conference on Artificial Intelligence.

In this study we aim to explore automatic methods that can detect online documents of low credibility, especially fake news, based on the style they are written in. We show that general-purpose text classifiers, despite seemingly good performance when evaluated simplistically, in fact overfit to sources of documents in training data. In order to achieve a truly style-based prediction, we gather a corpus of 103,219 documents from 223 online sources labelled by media experts, devise realistic evaluation scenarios and design two new classifiers: a neural network and a model based on stylometric features. The evaluation shows that the proposed classifiers maintain high accuracy in case of documents on previously unseen topics (e.g. new events) and from previously unseen sources (e.g. emerging news websites). An analysis of the stylometric model indicates it indeed focuses on sensational and affective vocabulary, known to be typical for fake news.

3rd International Summer School on Deep Learning

DeepLearn2019 logo 

Institute of Computer Science PAS is a coorganiyer of the 3rd International Summer School on Deep Learning "DeepLearn 2019". Summer School will take place on Julz 22-26, 2019 in Warsaw.

The summer school covers the issues of deep learning.

Summer school website:

XLIII Conference on Mathematical Statistics 2017


ICSPAS is a co-ogranizer of XLIII Conference on Mathematical Statistics 2017.
The conference will take place in Będlewo, Poland. The conference will start on 4/12/2017 and it will end on 8/12/2017.

The conference covers the issues of mathematical statistics and related applications.

Conference website:

Professora Zdzislaw Pawlak Award for Outstaning Monograth in Computer Science


The monograph entitled Algorytmy analizy skupień (en: Algorithms for Cluster Analysis) by Professors Sławomir T. Wierzchoń and Mieczysław A. Kłopotek has been honoured with Professor Zdzislaw Pawlak Award for Outstanding Monograph in Computer Science. The Award is granted by the Computer Science Committee of the Polish Academy of Sciences.

The monograph provides an overview of the basic algorithms for cluster analysis, together with their motivations and discussion of their various modifications. For example, the classic k-means algorithm, intended to group observations from a mixture of multivariate normal distributions with different mean vectors and similar covariance matrices is presented and a number of embodiments thereof are discussed, like the harmonic k-means, spherical k-means (used to document the analysis), and kernel variants for grouping the data that are not linearly separable. Not only the crisp partitions, but also fuzzy ones are considered. The latter allow for an alternative formulation of the k-means algorithm and its various variants (including kernel fuzzy algorithms, or spherical fuzzy algorithm) and extensions obtained through the introduction of the Dempster-Shafer mass function. This leads to so-called creedal partitions, for which - depending on the quality of the data - you can talk about a crisp, fuzzy, probabilistic, possibilistic or rough partitions reflecting various shades of data (un-)certainty.

In this context the problems of appropriate choice of parameters for these algorithms as well as of their efficient implementations are discussed. An attempt is made to systematize the methods of assessing the quality of the clusters. Quality measures of covering a given set of observations are also presented, as from a practical point of view, it is more natural to formulate an algorithm that generates coverage of a given set of objects, and not its partition.

A further contribution of the monograph is a unified exposition of spectral clustering and diffusive data analysis methods. Both classical spectral clustering algorithms and their modifications, resulting from various understandings of the term “cluster”, are presented. The relationship between graph cuts and selected spectral properties of graph Laplacians are investigated.

Formal analogies between various clustering methods are pointed out in order to show the possibilities of transferring modifications between methodologies for purposes of further algorithm development.

The monograph covers also specialisation of selected algorithms for tasks of processing very large data sets. A number of solutions to the problem of semi-supervised clustering are also presented.

An abbreviated version of the monograph in English is available from the Institute of Computer Science of Polish Academy of Sciences.

ATTENTION! This site uses cookies and similar technologies.

If you do not change your browser settings, you automatically agree to this.

I understand