The paper is concerned with the issues of modeling dynamic systems with interval parameters. In previous works, the authors proposed an adaptive interpolation algorithm for solving interval problems; the essence of the algorithm is the dynamic construction of a piecewise polynomial function that interpolates the solution of the problem with a given accuracy. The main problem of applying the algorithm is related to the curse of dimension, i.e., exponential complexity relative to the number of interval uncertainties in parameters. The main objective of this work is to apply the previously proposed adaptive interpolation algorithm to dynamic systems with a large number of interval parameters. In order to reduce the computational complexity of the algorithm, the authors propose using adaptive sparse grids. This article introduces a novelty approach of applying sparse grids to problems with interval uncertainties. The efficiency of the proposed approach has been demonstrated on representative interval problems of nonlinear dynamics and computational materials science.
We present an approach to solving parametric identification problems for dynamical systems. The approach is aimed at finding an interval estimate of the model parameters in which the solution of the corresponding modeling problem would contain the initial (experimental) data or minimize the deviation from them. The approach is based on an earlier-developed, tested, and justified adaptive interpolation algorithm for modeling dynamical systems with interval parameters. The problem of minimizing the distance between the solution of the interval problem and the experimental values of the phase variables in the space of bounds of the interval estimates of the model parameters is formulated. An expression for the gradient of the objective function to be minimized is obtained for further application of first-order optimization methods. The proposed approach is tested on a representative set of problems.
We consider the theoretical aspects of generalization of the adaptive interpolation algorithm to the case of a large number of interval uncertainties with the use of sparse grids. The classical adaptive interpolation algorithm essentially constructs an adaptive hierarchical partition of the uncertainty domain into subdomains each of which corresponds to a polynomial function of some degree interpolating the dependence of the solution of the problem on the point values of interval parameters with a given accuracy. The main disadvantage of the algorithm is its exponential dependence on the number of interval parameters. Already in the presence of five to six interval uncertainties, the algorithm becomes practically inapplicable. In this regard, it is proposed to use interpolation on adaptive sparse meshes instead of interpolation on a regular mesh; in some cases, this permits considerably expanding the scope of the algorithm. An estimate is produced for the global error of the algorithm on sparse meshes. The linear dependence of the global algorithm error on the mesh level is shown. The efficiency of our approach is demonstrated on a representative series of problems.
1 September 2021 in Computational Mathematics and Mathematical Physics
Problems with inexactly specified parameters arise in many applications, and they are often formulated as dynamical systems with interval parameters. For such systems, it is required to know how to obtain an interval estimate of solution given the interval values of the parameters. The main idea underlying the adaptive interpolation algorithm is to construct an adaptive hierarchical grid based on the kd-tree over the set formed by the interval initial conditions and parameters of the problem. In this grid, each cell contains an interpolation grid. For each point in time, depending on the features of solution, the decomposition is adaptively rearranged. At each step, the algorithm produces a piecewise polynomial function that interpolates the dependence of the problem solution on the parameter values with prescribed accuracy. As the number of interval parameters grows, the number of nodes in the interpolation grid grows exponentially, which restricts the algorithm scope. To mitigate this drawback, it is proposed to use the TT-decomposition of the multidimensional arrays in which the values of nodes of the interpolation grids are stored instead of these arrays. The efficiency of the proposed modification of the algorithm is demonstrated on model problems. In particular, the combustion of hydrogen–oxygen mixture in the presence of indeterminacy in the constants of chemical reaction rates is simulated.
The parametric identification problem for dynamical systems with rectangular and ellipsoid parameter uncertainty domains is solved for the case in which the experimental data are given in the form of intervals. The state of the considered dynamical systems at each moment of time is a parametric set. An objective function that characterizes the degree of deviation of the parametric sets of states from experimental interval estimates is constructed in the space of parameter uncertainty domains. To minimize the objective function, a sliding window algorithm has been developed, which is related to gradient methods. It is based on an adaptive interpolation algorithm that allows one to explicitly obtain parametric sets of states of a dynamical system within a given parameter uncertainty domain (window). The efficiency and performance of the proposed algorithm are demonstrated.
In this paper, we consider the issues of modeling an analog self-learning pulsed neural network based on memristive elements. One of the important properties of a memristor is the stochastic switching dynamics, which is mainly due to the stochastic processes of generation/recombination and the movement of ions (or oxygen vacancies) in a dielectric film under the effect of an electric field. The stochastic features are taken into account by adding a term responsible for the additive (Gaussian) noise to the memristor-state equation. The effect of noise on the functioning of an element is demonstrated using the dynamic model of a memristor as an example. The switching of a memristor from the high-resistance state to low-resistance state and vice versa is shown to occur from cycle to cycle in different ways, which is consistent with experimental data. We formulate a stochastic model that describes the hardware analog implementation of a pulsed neural network with memristive elements as synaptic weights and a learning mechanism based on the spike timing dependent plasticity (STDP) method. The operation of two neural networks consisting of one neuron with 64 synapses and two neurons with 128 synapses, respectively, is modeled. The recognition of 8 × 8 images is performed. The stochastic component in the memristor model is shown to have an effect on the fact that the templates from implementation to implementation are distributed among the neurons in different ways and the adaptation of weights (network training) occurs at different rates. In both cases, the network successfully learns to recognize the specified images.