Event Annoucements

Talk by Caroline Parker :

18.12.2024 15:00

Zajímá Vás multidisciplinární spolupráce mezi technologiemi, uměním a humanitními vědami?

Přijdte ve středu 18. prosince v 15:00 do CIIRC ČVUT na setkání s Caroline Parker, ředitelkou Creative z Edinburgh Futures Institute, které společně pořádají Centrum TAH a projekt RICAIP.

Caroline nás provede tématy, která otevírají nové příležitosti:

Creative Informatics: Modely spolupráce a řada praktických příkladů
Virtuální produkce: Od herního průmyslu po AI a film
Design thinking: Nástroj propojení akademie a průmyslu

Registrujte se co nejdřív!

Conference: Vision and Sports Summer School 2024

27.07.2024 - 27.07.2024

Let us invite you to the VS3 2024 Saturday Workshop that is jointly organized by the Visual Recognition Group and the Department of Cybernetics at the Czech Technical University in Prague, and by the Czech Pattern Recognition Society (CPRS), supported by the ELLIS Society and ELISE Consortium.

The list of talks:

Herve Jegou: Watermarking meets foundation models
Dima Damen: Opportunities in Egocentric Vision
Vittorio Ferrari: 3D Deep Learning
David Novotny: Reconstructing and Generating 3D Assets and Pose
Jiri Matas: Pixel-Level Out-of-Distribution Detection

Talk by David Cerna: Predicate invention in Popper - When what is known is not enough

26.06.2024 15:00

This talk will introduce Predicate Invention (PI) within inductive logic programming, explain why it is useful, and explain what types of predicate invention exist. This introduction is followed by a discussion of PI in Popper and David's work on extending the system by higher-order definitions and negated invented predicates. David will also cover current issues with PI and its use in applications. Speaker's academic bio: David Cerna is currently a Tenure-Track Researcher at the Czech Academy of Sciences Institute of Computer Science in the Department of Artificial Intelligence. He received his PhD in computer science from the Technical University of Vienna, and from 2015 to 2020, he worked as a postdoc on several projects at Johannes Kepler University. His research interests range from unification theory, automated reasoning, inductive synthesis, and proof theory.

Passcode: 671055

Talk by Josef Kittler: Information theoretic approach to deep learning

11.06.2024 14:00

Information divergence, which measures the distance between two probability distributions, has been the workhorse underpinning most of the objective functions for training deep neural networks advocated in the literature. We argue that this information measure is unable to pick the most robust solution from a potential set of candidates, which may impact on its generalisation properties. To address this problem, we propose an entropy regularised divergence as a loss function for training. We will show that this leads to a solution akin to multiple classifier fusion. The intricacies of its implementation will be discussed, as well as its relationship to contrastive learning in general, and the Nearest Proxy Triplet learning in particular.

Přednáška proběhne před valnou hromadou ČSKI ve 14 hod. 11.6.2024, v budově ČVUT, CIIRC, (Jug. partyzánů 3, Praha 6 - Dejvice), Červená posluchárna (vstup do budovy B, doprava po skleněném schodišti, první dveře vlevo (2. patro), po směrových šipkách bude možné využít i výtah).