Search for models

Stochastic Quasi-Gradient (SQG) methods have been developed for solving general optimization problems without exact calculation of objective function and constraints (let alone of their derivatives). SQG methods enable a sequential revision of approximate solutions towards the optimal using newly acquired information on the system, obtained via either direct on-line observations or(and) simulations. 

The linkage algorithms solve the problem of linking models, e.g. sectorial and/or regional, into an inter-sectorial inter-regional integrated model. Linkage enables to avoid “hard linking” of models in a single code, which saves the programming time and enables parallel distributed computations of individual models instead of a large scale integrated model. Models linkage preserves the structure of the original models taking into account critically important details, which are usually missing in aggregate models.

The current state of the world affairs calls for a revival of systems thinking to improve decision-making. Recognizing that the tightening of socio-economic links heightens the need for holistic responses, that disciplinary and sectorial solutions are of limited effectiveness and efficiency, and that big data is not generating integrative perspectives by itself, highlights the need for policymakers to become thoroughly familiar with the promises and pitfalls of systems analysis. Challenges are systemic, dynamic, and interconnected, and systems analysis, coupled with an improved anticipation, provides a coherent methodology and necessary tools to develop new approaches so urgently required for more coherent and effective policy planning.

We are developing and applying a range of methods for integrated multi-attribute evaluation under risk, subject to incomplete or imperfect information, and evaluations of decision situations using imprecise utilities, probabilities, and weights, as well as qualitative estimates between these components derived from sets of weight, utility and probability measures. To avoid some mathematical aggregation problems when handling set membership functions and similar, we use higher-order distributions for better discrimination between the possible outcomes.