PLAN RADA SEMINARA ZA JUN 2023. GODINE
Predavanja na seminaru mogu se pratiti na daljinu preko linka
https://miteam.mi.sanu.ac.rs/asset/YoqHWKALRkRTbK9So.
Registracija za on-line praćenje predavanja na Seminaru je na linku
https://miteam.mi.sanu.ac.rs/asset/xzGqvSp7aWbg8WpYX.
Utorak, 06.06.2023. u 14:15, Svečana sala Ogranka SANU u Novom Sadu i
Online
Neven Vrček, University of Zagreb, Faculty of Organization and Informatics, Croatia
SMART FRAMEWORKS FOR SMART INDUSTRY - SWARM AS A RESEARCH PLATFORM
Rapid technological development and convergence of technologies brought up a concept of smartness as a paradigm for modern society. We are talking about smart products, smart buildings, smart cities, smart ecosystems, etc. In order to fulfill expectations these concepts must be researched from many perspectives which overarch disciplines and reduce friction for widespread adoption of smart paradigm. Besides the obvious technological domain, we have to include legal aspects, user behavior, business models, etc. One approach to research these heterogeneous, and yet convergent domains, is to set up frameworks which can be used in various contexts and analyze them from different perspectives. This presentation introduces the SWARM of drones as a platform used to analyze various aspects that can adapt to the different domain requirements, most often within a smart urban environment. The platform developed within a project UNMANNED AERIAL VEHICLE CONTROL AND SURVEILLANCE FRAMEWORK is a testbed for the use and control of SWARM of drones. The platform combines technology-related research activities such as artificial intelligence for SWARM control, RF signal propagation and location determination within various use cases such as smart logistics, pollution identification, air quality monitoring etc. All these are accompanied with research about possible business models, legal requirements, privacy assessment, risk analysis and cost benefit analysis. The platform has interoperable modules that can be easily changed and adapted to various use cases. Our experience shows that such frameworks can gather scientists from various disciplines and bring up innovative solutions which have scientific and applied value.
Utorak, 13.06.2023. u 14:15, Knez Mihailova 36, sala 301f i
Online
Milan Kilibarda, Građevinski fakultet, Univerzitet u Beogradu
RFSI METOD ZA PROSTORNU INTERPOLACIJU
Uobičajeni problem u geoinformatici, geonaukama, zaštiti životne sredine tiče se prostorne interpolacije promenljivih iz tačkastih merenja, opažanja. Primeri takvih promenljivih su padavine, temperatura, kvalitet vazduha, zagađenje zemljišta i vode i drugi. Istorija prostorne interpolacije seže u rane dane geoinformatike, kada su razvijeni osnovni algoritmi kao što su najbliži sused i metod inverznih distanci. Geostatistika je napravila korak dalje razvijajući kriging interpolaciju, koja interpoluje merenja u prostoru u skladu sa stepenom njihove prostorne korelacije i ima dodatnu prednost što kvantifikuje interpolacionu grešku. Tokom godina razvijeno je mnogo različitih algoritama za kriging. Trenutno je regresioni kriging veoma popularan. Regresioni kriging obično proizvodi tačnije karte od običnog kriginga, jer pored informacija dobijenih iz merenja obližnjih tačaka, takođe koristi informacija sadržane u objašnjavajućim promenljivima o životnoj sredini (prediktori ili kovarijati). Kontinualni prediktori su na primer satelitski snimci, digitalni model terena i drugi rasterski podaci za koje imamo vrednost i na mestima gde je opažano i na mestima u kojim se računa vrednost interpolacije. Takođe su napravljena proširenja regresionog kriginga na prostorno-vremensku interpolaciju. Metode mašinskog učenja, kao što su: RF, SVM, ANN, GBM, Cubist nedavno su postale ozbiljan konkurent krigingu. Ove metode se oslanjaju isključivo na odnos između opažanja i prediktora i veoma su fleksibilne u svojoj strukturi modela. Ove metode mogu značajno poboljšati tačnost interpolacije, ako je dostupna dovoljna količina kontinualnih podataka za kreiranje modela i dovoljna količina opažanja za treniranje modela. Ipak, često za mnoge fenomene, količina dostupnih prediktora nije dovoljna da se napravi kvalitetan model za koji nije neophodno uključiti znanje o poziciji opažanja u proces interpolacije. Nedavno je publikovan rad baziran na Random Forest metodi koji se bavi prilagođavanjem modela za potrebe prostorne interpolacije, Random Forest Spatial Interpolation, RFSI. Ova tehnika zatvara krug prostorne interpolacije, ponovnim razmatranjem originalnih ideja prostorne interpolacije (npr. kriginga) i inkorporira merenja sa susednih lokacija u skup prediktora Random Forest modela.
U predavanju će biti dat pregled osnovnih struktura podataka koji se koriste u geoinformatici, pregled klasičnih metoda prostorne interpolacije, potom, predstavljena RFSI metoda i rezultati istraživanja primenjenih na modeliranje nekoliko prostornih fenomena, zatim, sledi kratak osvrt na RFSI inplementaciju u R softverskom okruženju u paketu meteo.
Predavanje je zasnovano na radu: Aleksandar Sekulić, Milan Kilibarda, Gerard B.M. Heuvelink, Mladen Nikolić, Branislav Bajat,
Random Forest Spatial Interpolation, Remote Sensing
1210 (2020), 1687.
https://doi.org/10.3390/rs12101687
Utorak, 20.06.2023. u 14:15, Knez Mihailova 36, sala 301f i
Online
Milica Knežević, Milan Todorović, Matematički institut SANU
PREDSTAVLJANJE PROJEKTA BC4GRID
BC4GRID je deo SMART4ALL projekta finansiranog u okviru Horizon 2020 programa. Matematički institut SANU je jedan od partnera na BC4GRID projektu čiji je fokus na primeni blokčejn tehnologije u pametnim elektroenergetskim mrežama (eng. smart grid). Mnogobrojna istraživanja pokazuju da blokčejn tehnologija ima značajan uticaj i sve veću upotrebu u različitim industrijskim sektorima. Razvoj energetskog sektora ide u smeru decentralizacije i efikasnije distribucije energije iz obnovljivih izvora, a to pruža motivaciju za uključivanje blokčejn tehnologije sa ciljem da se obezbedi pozdanost i efikasnost sistema. U okviru BC4GRID projekta biće razvijano blokčejn rešenje za optimizaciju postojećih procesa i trgovanje električnom energijom.
Zajednički sastanak sa seminarom Odlučivanje - teorija, thenologije i praksa.
Utorak, 27.06.2023. u 14:15, Knez Mihailova 36, sala 301f i
Online
Violeta Cvetkoska, Ss. Cyril and Methodius University in Skopje, Faculty of Economics – Skopje, North Macedonia
MEASURING ORGANIZATIONAL ABILITY IN THE WORLD OF FINANCE: UNVEILING INSIGHTS THROUGH DATA ENVELOPMENT ANALYSIS
Data Envelopment Analysis (DEA) has emerged as a prominent non-parametric methodology for measuring efficiency in various industries. In the field of finance, particularly in the banking sector, DEA has gained significant traction as a powerful tool for evaluating banking efficiency. Additionally, measurement of organizational ability is crucial for the banking sector as it enables informed decision-making, enhances operational efficiency, and promotes the overall stability and competitiveness of banks in an increasingly dynamic and demanding industry. During the lecture, we will delve into the practical implementation of DEA, showcasing how it can be utilized to measure and benchmark the organizational ability of banks. Additionally, we will explore the main drivers that influence this ability, shedding light on factors that contribute to or hinder the overall performance of banks in the finance sector. The findings presented in this lecture hold crucial implications for bank management, policymakers, and regulatory bodies, providing valuable insights to inform decision-making processes and strategic actions. Also, these findings serve as a catalyst for further research, encouraging banking professionals and researchers to conduct similar studies and contribute to the advancement of knowledge in the field.
RUKOVODIOCI SEMINARA
MI SANU
Vera Kovačević-Vujčić
Milan Dražić
FON
Zorica Bogdanovic
Marijana Despotovic-Zrakic
IEEE
Bozidar Radenkovic
