№ 2 (2024)

An overview of collaboration patterns is given. Common concepts are identified that collab oration pattern models, meta-models of pattern modelling languages, and ontology pattern models use in the pattern representations. A conceptual model for the usage of human-machine collaboration patterns in decision support is proposed. Options for the model usage are considered.

Some abilities of artificial intelligence methods application in predicting the outcomes of neurosurgical operations are discussed. The presented approach is based on the analysis of causal similarity as a basis for generation cause-and-effect dependencies initially hidden in accumulated empirical data. The mathematical formalization of this heuristic is constructed by clarifying similarity as a binary algebraic operation used to compare descriptions of precedents and search in them for approximate representation of the causality of target effects – the outcomes of neurosurgical operations. The possibilities of the presented approach are illustrated by the results of an intelligent analysis of real empirical data covering a series of neurosurgical operations of stem tumors performed in 2005-2018 at the N.N. Burdenko National Medical Research Center for Neurosurgery.

The article discusses normative systems (formal systems built genetically) of system-structural and system-object analysis. An improved alphabet of connections and nodes of the normative system of system-object analysis, based on the “Unit-Function-Object” approach, is proposed. Variants of meaningful interpretation of alphabetic nodes as a means for modeling various processes are pre- sented. For the process of manufacturing products from raw materials, the possibility of partial automation of the procedure for constructing graphic-analytical models of processes using an improved alphabet is shown.

The article proposes a mixed-integer multi-objective optimization algorithm based on the cuckoo search metaheuristic and the genetic crossover operator. Search in discrete space is carried out using a genetic operator, and in continuous space using a metaheuristic strategy. Performance was as- sessed using modified ZDT and DTLZ tests with mixed variables. The experimental results showed the high efficiency of the proposed algorithm on complex estimates of convergence and diversity.

Machine learning tasks focused on determining the laws of control of robots with complex locomotion are considered. The exponential computational complexity of such tasks is shown when using existing methods, in particular, reinforcement learning. The theoretical possibility of finding a multidimensional control function based on differential-algebraic equations of the dynamics of such systems is substantiated by varying the selected subset of the coupling equations. The possibility of a significant reduction in the dimension of the parameter space of the optimization problem on this basis is analyzed. Examples of the proposed method use for solving problems of the dynamics of machines, zoomorphic and anthropomorphic robots are given. The compatibility of the proposed mathematical method with neuromorphic dynamic systems used as a kernel in reservoir computing is shown. The fundamental admissibility of designing hardware for implementing reservoir computing on this basis is shown.

The paper considers a prototype of a three-dimensional convolutional neural network with an attention block detecting the probability of intracranial cerebral aneurysms in a single contrast computed tomography-angiography study. DICOM contrast computed tomography-angiography data with and without intracranial aneurysms were used to train the network. Metadata from the studies were not used. The data were divided into training and validation subsets in the proportion of 65% and 35%, respectively. Using Keras and Tensorflow libraries in the Python programming environment, a 192x192x128 three-dimensional convolutional neural network model with 4 convolutional layers, a kernel of dimension 3 and self-attention block was developed. The accuracy, precision and recall of classification on test samples reached 96%, 99% and 93% respectively that exceeded the performance of previously known neural networks.

A method is proposed for the fast selection of the blurriness coefficient of the kernel functions of the regression estimation of the probability density of a one-dimensional random variable. For a fast selection, the results of studying the asymptotic properties of the regression estimate of the probability density are used. A method for estimating the components of the optimal blurriness coefficient is proposed. The method of computational experiment is used to analyze the effectiveness of the proposed approach for a fast selection of the blurriness coefficient of the regression estimate of the probability density for a family of lognormal distribution laws for different volumes of initial data, and promising procedures for sampling the range of values of a random variable.