Proceedings of 2020 XXIII International Conference on Soft Computing and Measurements (SCM)
The Conference is traditionally focused on the actual problems in the field of Soft Computing and Measurement. Twenty third SCM Conferences, which took place in remote format in Zoom conference room of ETU “LETI”, during the previous years revealed great interest of both Russian and international researchers in this topic. Organizing and hosting the 2020 XXIII International Conference on Soft Computing and Measurements (SCM) in Russia is of great value for exchange of research ideas and practical results in this field, for discovering new problems and development trends, for development of new effective information systems targeted on solving complex practical problems. During the SCM-2020 Conference sessions, it is expected and planned to discuss a wide range of issues: uncertainty in measurements and computations, probabilistic methods of information processing, modeling of systems, control of complex objects under uncertainty, neural computing networks, genetic algorithms and their applications, methods and tools for development of expert and decision-making support systems, intellectual measuring systems, new approaches in measurement – fuzzy and soft measurement.
A challenging problem which arises in the domain of integrating symbolic and sub-symbolic computations within a massively parallel computational environments like Internet of Things is considered in application to the Linguistic Decision Making tasks. A novel theoretical idea is proposed on expressing linguistic operators in dynamics of an artificial neural network. The proposal consists of two consequent stages: expressing linguistic operators as structural manipulations and translating them in a neuroalgorithm. The theoretical foundation is Tensor Product Representation (TPR) that provides a generic framework of designing a neural network that does not require training and produces an exact result equivalent to the result of symbolic algorithms. This paper discusses viability of the proposed idea, demonstrates design of TPR-based arithmetic as a basic building block for construction of such a method and elaborates directions of further research.
There are multiple Multi-Attribute Decision Making methods elaborated for the past years. Those methods are targeted at aggregating assessments provided by the stakeholders of the problematic situation in order to choose the best alternative from the set of given ones. This paper considers ELECTRE, TOPSIS and the Multi-Level Linguistic Decision Making Methodology. In this paper we try to challenge first two methods by comparing it to latter method. One of the biggest challenges of modern Decision Making methods is flexibility to accept not only quantitative assessments but also hybrid ones: qualitative, interval, mixed etc. This brings the necessity for fuzzy computations. Decision Making methods analysis is performed through deep dive in constitution of each method and comparison across the set of elaborated criteria. Key criteria for the Decision Making methods assessment were identified and the comparative analysis on the base of two scenarios of different complexity was elaborated. The conclusion is made that ELECTRE and TOPSIS are well suited for small problems containing only several (less than a dozen) alternatives and criteria while being hardly generalized for the case of poorly structured problems (pollution, hunger, poverty). At the same time, Multi-Level Linguistic Decision Making Methodology excels at analyzing the problem from multiple aspects and considering any number of experts with arbitrary expertise that is beneficial in complex decision making cases.