Seminar for
DECISION MAKING – THEORY, TECHNOLOGY AND PRACTICE
PROGRAM
Predavanja možete pratiti i online putem MITEAM stranice Seminara Odlučivanje - teorija, tehnologija, praksa:
https://miteam.mi.sanu.ac.rs/asset/sEL32w8mjmruyeEqW
Plan rada Seminara Odlučivanje - teorija, tehnologija, praksa za MAJ 2025.
Četvrtak, 05.06.2025. u 13:00, Online
Darko Božanić, Military Academy, University of Defence in Belgrade
FUZZIFICATION OF THE ANALYTIC HIERARCHY PROCESS (AHP) METHOD USING A FUZZY SAATY SCALE WITH ADAPTIVE CONFIDENCE INTERVAL
The Analytic Hierarchy Process (AHP) method is one of the most widely used multi-criteria decision-making methods. In addition to the development of many other methods, the AHP method is still widely applied. This method is often modified by applying various mathematical tools, considering uncertainty and indeterminacy. This presentation will show an approach to modifying the AHP method using triangular fuzzy numbers with an adaptive confidence interval. In this approach, the confidence interval of the fuzzy number describing the comparison of criteria differs from one comparison to another. It depends on the opinion of the decision makers/experts, respectively, on their certainty in the comparison they make. In other words, the confidence interval of fuzzy numbers is variable and adjusts to the degree of confidence of the decision makers/experts in their claims. This approach allows the classic Saaty scale to be used when the experts are sure of their claims and the fuzzy Saaty scale to be used when they are less sure of their claims. It is particularly favorable for group decision-making.
Četvrtak, 12.06.2025. u 13:00, Pariske Komune bb, Niš i
Online
Sanja Stevanović, Mathematical Institute of the Serbian Academy of Sciences and Arts
PARALLEL MONTE CARLO-BASED SURROGATE OPTIMIZATION OF BUILDING ENERGY MODELS
Surrogate optimization optimizes expensive black-box functions with minimal number of evaluations. The aim is to quickly identify promising regions of the design space, usually without time constraints on selecting the next evaluation point. With EnergyPlus now supporting multi-threading, multiple building energy simulations can run in parallel. We present a simple method for parallel surrogate optimization of building energy models using XGBoost, which handles real, integer, and categorical variables. The approach uses Monte Carlo optimization to approximate a Pareto front between predicted performance and uncertainty. Candidates are grouped by uncertainty, and sampling combines top predictions, Pareto samples, and most uncertain points. We demonstrate the method on synthetic functions and a Kuwaiti villa model, showing better performance than popular surrogate and genetic optimization techniques.
Četvrtak, 19.06.2025. u 13:00, Pariske Komune bb, Niš i
Online
Lazar Velimirović, Mathematical Institute of the Serbian Academy of Sciences and Arts
MARKOV LOGIC NETWORKS
Markov Logic Networks (MLNs) represent a framework to reason under uncertainty in complex structured domains that combine first-order logic and probabilistic graphical models. This lecture will cover the foundation of MLNs, including the combination of probabilistic and logical reasoning, the structure of an MLN, and methods for inference and learning. The audience will see how probabilistic weights are used in MLNs to allow soft constraints and hear about a range of applications that use MLNs in real-world domains such as decision support systems. The lecture will show how MLNs can bridge the gap between statistical methods and symbolic AI.
Četvrtak, 26.06.2025. u 13:00,
Online
Ivana Štajner-Papuga, Faculty of Sciences, University of Novi Sad
FUZZY MEASURES IN THE DECISION-MAKING GALAXY
It is widely recognized that imprecision and uncertainties are unavoidable features of the real life, and the decision-making theory is a contemporary researched field that includes mathematical tools capable of modeling such real life problems. Fuzzy measure, introduced by Choquet in 1953 as capacity and, independently, by Sugeno in 1974 as fuzzy measure, offers a modern and powerful mathematical framework for addressing these complexities. The primary objective of this presentation is to highlight the role of fuzzy measures, i.e., monotone set functions, and corresponding integrals in the decision-making process. Unlike classical measures, fuzzy measures relax the condition of additivity and instead rely on monotonicity, making them more flexible for modeling the subtle nuances of unpredictability present in the human behavior.
Predavanja su namenjena sirokom krugu slusalaca, ukljucujuci studente redovnih i doktorskih studija. Seminar će se održavati svakog drugog četvrtka od 13:00 - 14:00h, CIITLAB, Elektronski fakultet Niš, Aleksandra Medvedeva 14, Niš
dr Lazar Velimirović
Rukovodilac seminara
dr Petar Vranić
Sekretar seminara
