Search academic texts

Information × Registration Number 0225U000160, (0124U002504) , R & D reports Title To develop generalized optimization methods for intelligent data analysis and their infrastructures popup.stage_title Розробити узагальнені методи оптимізації на частково впорядкованих множинах і структурованих даних Head Stovba Viktor O., Доктор філософії Registration Date 06-01-2025 Organization V.M. Glushkov Institute of Cybernetics of the National Academy of Sciences of Ukraine popup.description1 The goal of scientific research is to develop of generalized optimization methods on partially ordered sets and structured data popup.description2 For finding parameters of linear regression model with L1-regularization and criterion of moduli sum of residuals powered to p∈[1,2], the emlmpr algorithm is constructed. The results of computational experiments demonstrate program running time, robustness of the problem solutions at p≈1, and possibility of restoring parameters of linear regression model if there is a linear dependence between the model factors using L1-regularization. For two-stage transportation problem with two-sided constraints on consumer demands and upper bounds on throughput of intermediate points, necessary and sufficient conditions for constraint system consistency are substantiated. The considered two-stage transportation problem is applied to the model problem of optimal partitioning-allocation. Results of six computational experiments on solving this problem using the CPLEX solver with different discretization, upper bounds on throughput of intermediate points, production and consumption volumes are presented. For the chosen discretization, the problem has 1.25 million variables and 250 thousand linear constraints. Transportation plans for each experiment and corresponding optimal set partitions are presented. Models and methods of group decision-making are investigated and their application to generalized group cybernetic systems, which may consist of a manager and a subordinate, customer and performer, state regulator and private enterprise, etc. Problems that arise when attempting to extend the analysis of individual decisions to group decisions are considered, in particular, Arrow's paradox when combining ranking of alternatives of individual group members to obtain a (cumulative) ranking of the entire group, and the prisoner's dilemma are considered. Mixed strategies of cooperation and generalized leadership are presented, as well as main theorems that characterize the behavior of participants in these conditions and desired characteristics, in particular their output and profit. Product Description popup.authors Dunaievskyi Maksym S. Suleimanov Seit-Bekir Sinaver Ogly popup.nrat_date 2025-01-06 Close