All issues

Stochastic optimization is a current area of research due to significant advances in machine learning and their applications to everyday problems. In this paper, we consider two fundamentally different methods for solving the problem of stochastic optimization — online and offline algorithms. The corresponding algorithms have their qualitative advantages over each other. So, for offline algorithms, it is required to solve an auxiliary problem with high accuracy. However, this can be done in a distributed manner, and this opens up fundamental possibilities such as, for example, the construction of a dual problem. Despite this, both online and offline algorithms pursue a common goal — solving the stochastic optimization problem with a given accuracy. This is reflected in the comparison of the computational complexity of the described algorithms, which is demonstrated in this paper.

The comparison of the described methods is carried out for two types of stochastic problems — convex optimization and saddles. For problems of stochastic convex optimization, the existing solutions make it possible to compare online and offline algorithms in some detail. In particular, for strongly convex problems, the computational complexity of the algorithms is the same, and the condition of strong convexity can be weakened to the condition of $\gamma$-growth of the objective function. From this point of view, saddle point problems are much less studied. Nevertheless, existing solutions allow us to outline the main directions of research. Thus, significant progress has been made for bilinear saddle point problems using online algorithms. Offline algorithms are represented by just one study. In this paper, this example demonstrates the similarity of both algorithms with convex optimization. The issue of the accuracy of solving the auxiliary problem for saddles was also worked out. On the other hand, the saddle point problem of stochastic optimization generalizes the convex one, that is, it is its logical continuation. This is manifested in the fact that existing results from convex optimization can be transferred to saddles. In this paper, such a transfer is carried out for the results of the online algorithm in the convex case, when the objective function satisfies the $\gamma$-growth condition.

Modeling combat operations is an urgent scientific and practical task aimed at providing commanders and staffs with quantitative grounds for making decisions. The authors proposed the function of victory in combat and military operations, based on the function of the conflict by G. Tullock and taking into account the scale of combat (military) operations. On a sufficient volume of military statistics, the scale parameter was assessed and its values were found for the tactical, operational and strategic levels. The game-theoretic models «offensive – defense», in which the sides solve the immediate and subsequent tasks, having the formation of troops in one or several echelons, have been investigated. At the first stage of modeling, the solution of the immediate task is found — the breakthrough (holding) of defense points, at the second — the solution of the subsequent task — the defeat of the enemy in the depth of the defense (counterattack and restoration of defense). For the tactical level, using the Nash equilibrium, solutions were found for the closest problem (distribution of the forces of the sides by points of defense) in an antagonistic game according to three criteria: a) breakthrough of the weakest point, b) breakthrough of at least one point, and c) weighted average probability. It is shown that it is advisable for the attacking side to use the criterion of «breaking through at least one point», in which, all other things being equal, the maximum probability of breaking through the points of defense is ensured. At the second stage of modeling for a particular case (the sides are guided by the criterion of breaking through the weakest point when breaking through and holding defense points), the problem of distributing forces and facilities between tactical tasks (echelons) was solved according to two criteria: a) maximizing the probability of breaking through the defense point and the probability of defeating the enemy in depth defense, b) maximizing the minimum value of the named probabilities (the criterion of the guaranteed result). Awareness is an important aspect of combat operations. Several examples of reflexive games (games characterized by complex mutual awareness) and information management are considered. It is shown under what conditions information control increases the player’s payoff, and the optimal information control is found.

In our research, we focus on the problem of classification for silent speech recognition to develop a brain– computer interface (BCI) based on electroencephalographic (EEG) data, which will be capable of assisting people with mental and physical disabilities and expanding human capabilities in everyday life. Our previous research has shown that the silent pronouncing of some words results in almost identical distributions of electroencephalographic signal data. Such a phenomenon has a suppressive impact on the quality of neural network model behavior. This paper proposes a data processing technique that distinguishes between statistically remote and inseparable classes in the dataset. Applying the proposed approach helps us reach the goal of maximizing the semantic load of the dictionary used in BCI.

Furthermore, we propose the existence of a statistical predictive criterion for the accuracy of binary classification of the words in a dictionary. Such a criterion aims to estimate the lower and the upper bounds of classifiers’ behavior only by measuring quantitative statistical properties of the data (in particular, using the Kolmogorov – Smirnov method). We show that higher levels of classification accuracy can be achieved by means of applying the proposed predictive criterion, making it possible to form an optimized dictionary in terms of semantic load for the EEG-based BCIs. Furthermore, using such a dictionary as a training dataset for classification problems grants the statistical remoteness of the classes by taking into account the semantic and phonetic properties of the corresponding words and improves the classification behavior of silent speech recognition models.

The problem of optimal control of traffic flow in an urban road network is considered. The control is carried out by varying the duration of the working phases of traffic lights at controlled intersections. A description of the control system developed is given. The control system enables the use of three types of control: open-loop, feedback and manual. In feedback control, road infrastructure detectors, video cameras, inductive loop and radar detectors are used to determine the quantitative characteristics of current traffic flow state. The quantitative characteristics of the traffic flows are fed into a mathematical model of the traffic flow, implemented in the computer environment of an automatic traffic flow control system, in order to determine the moments for switching the working phases of the traffic lights. The model is a system of finite-difference recurrent equations and describes the change in traffic flow on each road section at each time step, based on retrived data on traffic flow characteristics in the network, capacity of maneuvers and flow distribution through alternative maneuvers at intersections. The model has scaling and aggregation properties. The structure of the model depends on the structure of the graph of the controlled road network. The number of nodes in the graph is equal to the number of road sections in the considered network. The simulation of traffic flow changes in real time makes it possible to optimally determine the duration of traffic light operating phases and to provide traffic flow control with feedback based on its current state. The system of automatic collection and processing of input data for the model is presented. In order to model the states of traffic flow in the network and to solve the problem of optimal traffic flow control, the CTraf software package has been developed, a brief description of which is given in the paper. An example of the solution of the optimal control problem of traffic flows on the basis of real data in the road network of Moscow is given.

Optimal management of fuel supply system boils down to choosing an energy development strategy which provides consumers with the most efficient and reliable fuel and energy supply. As a part of the program on switching the heat supply distributed management system of the Udmurt Republic to renewable energy sources, an “Information-analytical system of regional alternative fuel supply management” was developed. The paper presents the mathematical model of optimal management of fuel supply logistic system consisting of three interconnected levels: raw material accumulation points, fuel preparation points and fuel consumption points, which are heat sources. In order to increase effective the performance of regional fuel supply system a modification of information-analytical system and extension of its set of functions using the methods of quick responding when emergency occurs are required. Emergencies which occur on any one of these levels demand the management of the whole system to reconfigure. The paper demonstrates models and algorithms of optimal management in case of emergency involving break down of such production links of logistic system as raw material accumulation points and fuel preparation points. In mathematical models, the target criterion is minimization of costs associated with the functioning of logistic system in case of emergency. The implementation of the developed algorithms is based on the usage of genetic optimization algorithms, which made it possible to obtain a more accurate solution in less time. The developed models and algorithms are integrated into the information-analytical system that enables to provide effective management of alternative fuel supply of the Udmurt Republic in case of emergency.

Simulation of combat and military operations is the most important scientific and practical task aimed at providing the command of quantitative bases for decision-making. The first models of combat were developed during the First World War (M. Osipov, F. Lanchester), and now they are widely used in connection with the massive introduction of automation tools. At the same time, the models of combat and war do not fully take into account the moral potentials of the parties to the conflict, which motivates and motivates the further development of models of battle and war. A probabilistic model of combat is considered, in which the parameter of combat superiority is determined through the parameter of moral (the ratio of the percentages of the losses sustained by the parties) and the parameter of technological superiority. To assess the latter, the following is taken into account: command experience (ability to organize coordinated actions), reconnaissance, fire and maneuverability capabilities of the parties and operational (combat) support capabilities. A game-based offensive-defense model has been developed, taking into account the actions of the first and second echelons (reserves) of the parties. The target function of the attackers in the model is the product of the probability of a breakthrough by the first echelon of one of the defense points by the probability of the second echelon of the counterattack repelling the reserve of the defenders. Solved the private task of managing the breakthrough of defense points and found the optimal distribution of combat units between the trains. The share of troops allocated by the parties to the second echelon (reserve) increases with an increase in the value of the aggregate combat superiority parameter of those advancing and decreases with an increase in the value of the combat superiority parameter when repelling a counterattack. When planning a battle (battles, operations) and the distribution of its troops between echelons, it is important to know not the exact number of enemy troops, but their capabilities and capabilities, as well as the degree of preparedness of the defense, which does not contradict the experience of warfare. Depending on the conditions of the situation, the goal of an offensive may be to defeat the enemy, quickly capture an important area in the depth of the enemy’s defense, minimize their losses, etc. For scaling the offensive-defense model for targets, the dependencies of the losses and the onset rate on the initial ratio of the combat potentials of the parties were found. The influence of social costs on the course and outcome of wars is taken into account. A theoretical explanation is given of a loss in a military company with a technologically weak adversary and with a goal of war that is unclear to society. To account for the influence of psychological operations and information wars on the moral potential of individuals, a model of social and information influence was used.

In the presented work, a hybrid optimal simulation model of an assembly shop in the AnyLogic environment has been developed, which allows you to select the parameters of production systems. To build a hybrid model of the investigative approach, discrete-event modeling and aggressive modeling are combined into a single model with an integrating interaction. Within the framework of this work, a mechanism for the development of a production system consisting of several participants-agents is described. An obvious agent corresponds to a class in which a set of agent parameters is specified. In the simulation model, three main groups of operations performed sequentially were taken into account, and the logic for working with rejected sets was determined. The product assembly process is a process that occurs in a multi-phase open-loop system of redundant service with waiting. There are also signs of a closed system — scrap flows for reprocessing. When creating a distribution system in the segment, it is mandatory to use control over the execution of requests in a FIFO queue. For the functional assessment of the production system, the simulation model includes several functional functions that describe the number of finished products, the average time of preparation of products, the number and percentage of rejects, the simulation result for the study, as well as functional variables in which the calculated utilization factors will be used. A series of modeling experiments were carried out in order to study the behavior of the agents of the system in terms of the overall performance indicators of the production system. During the experiment, it was found that the indicator of the average preparation time of the product is greatly influenced by such parameters as: the average speed of the set of products, the average time to complete operations. At a given limitation interval, we managed to select a set of parameters that managed to achieve the largest possible operation of the assembly line. This experiment implements the basic principle of agent-based modeling — decentralized agents make a personal contribution and affect the operation of the entire simulated system as a whole. As a result of the experiments, thanks to the selection of a large set of parameters, it was possible to achieve high performance indicators of the assembly shop, namely: to increase the productivity indicator by 60%; reduce the average assembly time of products by 38%.

The paper describes a method for recognizing authors of literary texts by the proximity of fragments into which a separate text is divided to the standard of the author. The standard is the empirical frequency distribution of letter combinations, built on a training sample, which included expertly selected reliably known works of this author. A set of standards of different authors forms a library, within which the problem of identifying the author of an unknown text is solved. The proximity between texts is understood in the sense of the norm in L1 for the frequency vector of letter combinations, which is constructed for each fragment and for the text as a whole. The author of an unknown text is assigned the one whose standard is most often chosen as the closest for the set of fragments into which the text is divided. The length of the fragment is optimized based on the principle of the maximum difference in distances from fragments to standards in the problem of recognition of «friend–foe». The method was tested on the corpus of domestic and foreign (translated) authors. 1783 texts of 100 authors with a total volume of about 700 million characters were collected. In order to exclude the bias in the selection of authors, authors whose surnames began with the same letter were considered. In particular, for the letter L, the identification error was 12%. Along with a fairly high accuracy, this method has another important property: it allows you to estimate the probability that the standard of the author of the text in question is missing in the library. This probability can be estimated based on the results of the statistics of the nearest standards for small fragments of text. The paper also examines statistical digital portraits of writers: these are joint empirical distributions of the probability that a certain proportion of the text is identified at a given level of trust. The practical importance of these statistics is that the carriers of the corresponding distributions practically do not overlap for their own and other people’s standards, which makes it possible to recognize the reference distribution of letter combinations at a high level of confidence.

Distributed saddle point problems (SPPs) have numerous applications in optimization, matrix games and machine learning. For example, the training of generated adversarial networks is represented as a min-max optimization problem, and training regularized linear models can be reformulated as an SPP as well. This paper studies distributed nonsmooth SPPs with Lipschitz-continuous objective functions. The objective function is represented as a sum of several components that are distributed between groups of computational nodes. The nodes, or agents, exchange information through some communication network that may be centralized or decentralized. A centralized network has a universal information aggregator (a server, or master node) that directly communicates to each of the agents and therefore can coordinate the optimization process. In a decentralized network, all the nodes are equal, the server node is not present, and each agent only communicates to its immediate neighbors.

We assume that each of the nodes locally holds its objective and can compute its value at given points, i. e. has access to zero-order oracle. Zero-order information is used when the gradient of the function is costly, not possible to compute or when the function is not differentiable. For example, in reinforcement learning one needs to generate a trajectory to evaluate the current policy. This policy evaluation process can be interpreted as the computation of the function value. We propose an approach that uses a smoothing technique, i. e., applies a first-order method to the smoothed version of the initial function. It can be shown that the stochastic gradient of the smoothed function can be viewed as a random two-point gradient approximation of the initial function. Smoothing approaches have been studied for distributed zero-order minimization, and our paper generalizes the smoothing technique on SPPs.

Algorithms of wireless sensor networks operation based on modified ZigBee/IEEE 802.15.4 protocol stack and problems of energy saving with simultaneous decrease of network latency are studied. Theoretical computations are given. Roles distribution and routers schedule assignment algorithms are described. Both results of experiments carried out with real devices and results of simulations with ns-2 (open-source network simulator) are given and analyzed.