PLAN RADA SEMINARA ZA NOVEMBAR 2024. GODINE
Predavanja na seminaru mogu se pratiti na daljinu preko linka
https://miteam.mi.sanu.ac.rs/asset/wnz6oyxsQsy29LfJA.
Registracija za on-line praćenje predavanja na Seminaru je na linku
https://miteam.mi.sanu.ac.rs/asset/xzGqvSp7aWbg8WpYX.
Neulogovani korisnici mogu pratiti prenos predavanja na ovom linku (ali ne mogu postavljati pitanja osim putem chata):
https://miteam.mi.sanu.ac.rs/call/wnz6oyxsQsy29LfJA/MjQ__eH607WeAL9X7IFtUI98xdQQgVkp-ljiEKPPfXr.
Utorak, 05.11.2024. u 14:15, Kneza Mihaila 36, sala 301f i
Online
Arutyun Avetisyan, Andrey Belevantsev, Yuri Markin, Institute for System Programming of the Russian Academy of Sciences (ISP RAS)
DEVELOPMENT AND VERIFICATION OF ARTIFICIAL INTELLIGENCE TECHNOLOGY
The report will primarily focus on ISP RAS research for trusted AI technology, including trusted frameworks for training neural networks, the theory and practice of creating models that can withstand attacks, preventing model aging, searching for machine learning vulnerabilities, and others.
The second part of the report will be devoted to the development of secure software. The report will present technologies developed at ISP RAS for secure software development (SDLC), which are necessary for creating efficient and secure software of any type, including artificial intelligence. Among them are approaches to program analysis, including static, dynamic analysis and fuzzing, secure compilation methods, and attack detection methods.The final part of the report will be devoted to research on digital watermarks, which ISP RAS conducts jointly with the Steklov Mathematical Institute of the Russian Academy of Sciences. The rapid development of AI poses new challenges in protecting training datasets, as well as trained neural network models, from anonymous theft.
Therefore, on the one hand, it is necessary to guarantee the possibility of establishing the fact of content synthesis, and on the other hand, to prevent the creation of deepfakes based on it.
Zajednički sastanak sa seminarom Veštačka inteligencija.
Utorak, 12.11.2024. u 14:15, Online
Miloš Simić, Univerzitet u Beogradu
NOVE METODE ZA KONTROLU GREŠKE BINARNIH KLASIFIKATORA U MAŠINSKOM UČENJU
Postoje dve vrste grešaka koje mogu napraviti binarni klasifikatori. U pitanju su klasifikacija pozitivnog objekta kao negativnog (engl. false negative) i negativnog objekta kao pozitivnog (engl. false positive). U određenim oblastima, jedna od ove dve greške je kritična i nosi daleko ozbilјnije posledice, pa je cilј razviti klasifikatore kod kojih je njena verovatnoća manja od unapred definisanog praga. U ovom predavanju će biti predstavlјene metode koje su u okviru mašinskog učenja predložene za kontrolu verovatnoće greške binarnih klasifikatora. Pored njih, biće predstavlјene i nove tehnike kontrole zasnovane na statističkim testovima, kao i rezultati evaluacije na više skupova podataka.
Ponedeljak, 18.11.2024. u 16:00, Pariske Komune bb, Niš i
Online
Anđelka Hedrih, Mathematical Institute SANU, Belgrade, Serbia
BELOUSOV-ZHABOTINSKY SYSTEM AND ITS APPLICATION IN ROBOTICS
The Belousov-Zhabotinsky (BZ) system is one of the most famous examples of oscillating chemical reactions, where bromate ions oxidize malonic acid to produce carbon dioxide, with ferroin serving as a catalyst. This reaction is a remarkable example of a nonlinear chemical oscillator that exhibits self-organizing behavior across both spatial and temporal dimensions. Such nonlinearity allows the system to form complex, dynamic patterns, often likened to those observed in biological and physical systems. Over the years, various substances have been documented in scientific literature for their ability to modulate the dynamics of the BZ reaction. Examples include silver ions, heavy water, lipid bilayers, liposomes, electric potentials, light exposure, methanol, ethylene glycol, ascorbic acid, tannic acid, among others. Each of these modulators uniquely influences the oscillatory behavior of the reaction, potentially altering its frequency, amplitude, or spatial patterns. These modulatory effects of the BZ system have intriguing applications in fields such as robotics, where the reaction can serve as a chemical engine for small-scale robots. For instance, studies demonstrate the use of BZ-based systems in programmable hybrid digital-chemical processors, as well as in BZ liquid marbles, which act as miniature chemical reactors capable of supporting robotic motion and control. We will present and discuss these BZ applications in robotic systems and our recent research that uncovered new modulators of the BZ reaction, expanding the possibilities for precision control in both chemical processing and robotic applications.
Zajednički sastanak sa seminarima Odlučivanje - teorija, thenologije, praksa i Biomehanika, bioinžinjering i matematička biologija.
Utorak, 19.11.2024. u 14:15, Online
Siniša Tomović, Matematički institut SANU
METODE DOKAZIVANJA SIGURNOSTI RFID AUTENTIFIKACIONIH PROTOKOLA ZASNOVANIH NA LPN PROBLEMU
RFID tehnologija omogućava identifikaciju i autentifikaciju objekata putem radio talasa i postala je neizostavni deo svakodnevice savremenog čoveka, sa mnoštvom oblasti primene poput lanaca snabdevanja, transporta, kontrole pristupa, zdravstva, upravljanja inventarom i mnogih drugih. Otuda je od ključne važnosti osigurati efikasnost i bezbednost ovog vida autentifikacije, naročito u kritičnim i osetljivim slučajevima upotrebe. Na ovom predavanju prikazaćemo tehnike dokazivanja sigurnosti odabranih RFID autentifikacionih protokola, uključujući i protokol koji smo mi razvili - NHB#, redukcijom na poznati NP-kompletan LPN problem i njegove varijante.
Utorak, 26.11.2024. u 14:15, Online
Đorđe Jovanović, Mathematial Institute SANU
HONEY BADGER ALGORITHM FOR COMMUNITY DETECTION
The community detection problem is crucial in understanding the structure and dynamics of complex networks, such as social, biological, and technological systems. Identifying communities, or clusters of nodes with dense interconnections, helps reveal hidden patterns, optimize information flow, and understand group behavior. Since this is a NP-hard problem, an attempt to solve it using metaheuristics can be made. This presentation analyzes the performance of Honey Badger Optimization Algorithm (HBO for short) in solving the community detection problem. Its main focus is showing the necessary steps in adapting a continual optimization problem to a discrete one, design decisions used in adapting the HBO algorithm to community detection problem, and searching the best set of input parameter values.
RUKOVODIOCI SEMINARA
MI SANU
Vera Kovačević-Vujčić
Milan Dražić
FON
Zorica Bogdanovic
Marijana Despotovic-Zrakic
IEEE
Bozidar Radenkovic
