MISANU

PROGRAM

Plan rada Seminara iz veštačke inteligencije za OKTOBAR 2024.

Registraciona forma za učesće, i link na predavanje ako ste već registrovani:
https://miteam.mi.sanu.ac.rs/asset/CW5nJWDSEZDj7p32p
Ukoliko želite samo da gledate predavanje bez mogućnosti aktivnog učešća, prenos će biti dostupan na:
https://miteam.mi.sanu.ac.rs/asset/4LNW8WtML7rLKojoz
Na ovom linku se mogu pronaci kratka uputstva na srpskom i engleskom:
https://miteam.mi.sanu.ac.rs/asset/Kc7qJtEvoMFx9MFnz

Sreda, 02.10.2024. u 19:00, Online
Miloš Nikolić, School of Informatics, University of Edinburgh
EFFICIENT MACHINE LEARNING OVER RELATIONAL DATA
Relational data, a cornerstone of many data science applications, presents unique challenges for machine learning. Traditional approaches often involve extracting large, flat datasets from database systems and feeding them to specialised machine learning tools. This process discards valuable information about the structure of the training data, leading to an expensive learning phase, especially for large datasets. This talk explores recent advancements in training and maintaining machine learning models directly on relational data. By exploiting the underlying relational structure, we can improve the runtime performance of the learning task by several orders of magnitude.

Sreda, 09.10.2024. u 19:00, Online
Bela Stantic, Big Data and Smart Analytics lab, Griffith University, Gold Coast, Australia
CREATING HARMONY: ADVANCING AI AND ITS IMPACT ON RESEARCH
In this talk, Professor Bela Stantic will first elaborate on diverse funded projects at the Big Data and Smart Analytics lab, at Griffith University. Many of these projects rely on pre-trained-transformers, large language models, and Generative AI, therefore, a brief introduction on what they are and how they work will be provided. The presentation will include demonstrations of several pre-trained and fine-tuned large language models and GenerativeAI for several domains running locally on the Griffith Big Data cluster. Prof Stantic will introduce the architecture and demonstrate DORIS (Digital Oracle for Research in Sciences), a GenerativeAI system that he developed and purposely trained by trusted data sources to do diverse tasks including reviewing scientific papers. The talk will conclude by elaborating on the impact that the advancing AI is making on research and how harmony can be created despite such a fast pace of AI advancement in the last couple of years.

Sreda, 16.10.2024. u 19:00, Online
Michael Benedikt, Director of Advanced MSc in Computer Science
LOGIC AND ASYMPTOTIC COMBINATORICS OF GRAPH NEURAL NETWORKS
Graph neural networks (GNNs) are the predominant architectures for a variety of learning tasks on graphs. We present a new angle on the expressive power of GNNs by studying how the predictions of a GNN probabilistic classifier evolve as we apply the classifier on larger graphs drawn from some random graph model. We show that the output converges asymptotically almost surely to a constant function, which upper-bounds what these classifiers can express uniformly.
Our convergence results are framed within a query language with aggregates, subsuming a very wide class of GNNs, including state of the art models, with aggregates including mean and the attention-based mechanism of graph transformers. The results apply to a broad class of random graph models, but in the talk we will focus on Erdős-Rényi model and the stochastic block model. The query language-based approach allows our results to be situated within the long line of research on convergence laws for logic. The talk will include joint work with Sam Adam-Day, Ismail Ceylan, and Ben Finkeshtein - see https://arxiv.org/abs/2403.03880, and also joint work with Sam Adam-Day and Alberto Larrauri.

Utorak, 23.10.2024. u 14:15, Online
Dragan Jočić, Matematički institut SANU
USLOVNA DISTRIBUTIVNOST OPERATORA AGREGACIJE SA ABSORBUJUĆIM ELEMENTOM I ODGOVARAJUĆA FUNKCIJA KORISNOSTI
Operatori (funkcije) agregacije su važno matematičko sredstvo u različitim oblastima (inženjerstvo, ekonomija, informatika, medicina itd), a pre svega u raznim pristupima problemu odlučivanja. Pored minimalnih osobina koje svaki operator agregacije mora posedovati (monotonost i granični uslovi), u zavisnosti od konkretne primene ove funkcije mogu posedovati i još neke dodatne osobine: komutativnost, asosijativnost, neprekidnost, neutralni i abssorbujući element itd. Od posebnog interesa je proučavanje dodatnih osobina operatora agregacije koje su izvedene iz rešenja funkcionalnih jednačina koje uključuju ove operatore. Neke od tih osobina su migrativnost, modularnost, idempotentnost, distributivnost itd. U ovom predavanju predstavićemo neke poznate klase operatora agregacije sa absorbujućim elementom i razmotrićemo jedan poseban oblik jednačina distributivnosti (uslovnu distributivnost) za te operatore. Takođe, pokazaćemo primenu dobijemih distributivnih parova u konstrukciji funkcije korisnosti i analiziraćemo ponašanje donosioca odluke u odnosu na nju.

Sreda, 30.10.2024. u 19:00, Online
Zorica Bogdanović, Univerzitet u Beogradu, Fakultet organizacionih nauka
PRIMENA VEŠTAČKE INTELIGENCIJE U RAZVOJU SOFTVERA
Digitalna transformacija i brzo usvajanje modela i tehnologija elektronskog poslovanja u preduzećima zahtevaju potupno usklađivanje razvoja softvera i IT operacija. Ovo se postiže primenom DevOps principa, metoda i alata za kontinulani razvoj softvera, u svim fazama softverskog procesa, uključujući analizu, projektovanje, razvoj, integraciju, testiranje, isporuku i praćenje performansi. Cilj ovog predavanja je da se analiziraju mogućnosti primene veštačke inteligencije u navedenim fazama DevOps pristupa razvoju softvera. Posebno će biti razmotreni alati i najbolje prakse čija primena treba da omogući povećanje produktivnosti, bolje upravljanje resursima i veći kvalitet softvera.

Ovaj onlajn seminar nastao je kao nastavak sastanka "Serbian AI Meeting" i zamišljen je da na njemu istraživači iz Srbije i iz dijaspore, kao i istraživači sa univerzteta, naučnih instituta i iz prakse predstavljaju naučne teme i rezultate iz oblasti veštačke inteligencije.
Link za svako pojedinačno predavanje biće dostavljen dan pre održavanja predavanja.

Andreja Tepavčević
Rukovodilac seminara

Biljana Stojanović
Sekretar seminara