In order to participate in the seminar, on October 16 after 4:45 pm (UTC+7) you should connect to the Zoom conference via the following link https://zoom.us/j/93473879731 or manually using the Zoom conference ID 934 7387 9731.
One of the traditional tasks of artificial intelligence is generative design — the synthesis of formal descriptions of digital objects (including ones corresponding to the phenomena of the real world) that a priori satisfy a certain set of requirements, under uncertain conditions and incompleteness of the initial data, by means of evolutionary calculations. Classical applications of generative design are, for example, the design of actuators for robots, structures of buildings, as well as the determination of the molecular structure of a given matter in order to create functional materials. In all these cases, the key success factor is the correct formalization of the fitness-function, which makes it possible to rank the obtained solutions. In the presence of quality criteria based on a description of the phenomenon itself, it is limited only by the intensity of the computation.
However, the situation changes dramatically with the transition to the generative design of weakly formalized systems with a significant influence of the human factor. In this case, many characteristics can be categorical, reflecting the properties of the system at a qualitative level; however, the possibility (and even necessity) of their exact attainability is not always obvious. Therefore, for the generative design of such systems, an approach based on value modeling is proposed, which makes it possible to formalize the fitness-function in the proximity logics of the generated solution to a certain set of terminal values determined by the subject area. Solving the problem of quantifying values by means of multidimensional scaling allows you to transfer into a space in which the problem can be solved in terms of finding the optimal trajectory, which allows us to express the characteristics of the solution in terms of the resources (time, materials, etc.) spent on achieving it. The report will consider a general approach to modeling value-oriented systems, and also illustrate its application for generative design in tasks:
(a) optimal planning for the urban areas development,
(b) managing personal educational curricula,
(c) formation of loyalty programs,
(d) design of port facilities.
Special attention will be paid to the issues of generative design of the models themselves in analytical form and their numerical implementation.