All issues

Reliable plant disease diagnosis requires not only high classification accuracy, but also stable generalization and interpretable decision-making. Although attention mechanisms are useful to these deep learning models, the performance also depends on where and how they are integrated into the network architecture. This study presents a design-optimized YOLO11m-based classification framework that systematically investigates the impact of Convolutional Block Attention Module (CBAM) injection at different architectural levels for plant disease diagnosis. We perform a comparative modelcontrolled analysis of three model architectures: (i) the baseline YOLO11m-Cls architecture lacking attention, (ii) only adding the backbone block along with CBAM and (iii) a hybrid architecture that includes reduced backbone attention combined with CBAM added at classification head. All models are trained and tested under the same experimental settings using a largescale dataset with around 90 000 images of 38 types of plant diseases. Experimental results clearly show that the uniform injection of CBAM into the backbone reduces stability but causes generalization to worsen with a higher validation loss and significantly lower Top-1 accuracy (≈ 90.5%), while the hybrid attention design balances stability and discrimination, with Top-1 accuracy up to 99.71%, Top-5 accuracy up to 99.99% and near-baseline validation behavior respectively Grad-CAMbased interpretability analysis also demonstrates that the hybrid model generates enhanced and biologically interpretable activation maps, which are more disease-specific with less distraction from background. Notably, the aim of our work is not for achieving superior performance over all classifiers but instead only to provide design-level evidence on how placing attention modulates rigidity and interpretability in YOLO-based classification models. The results provide practical architectural considerations for building dependable and interpretable AI systems in agriculture.

In the article, the questions of application of a optimization method for support of planning processes in warehouse systems by means of simulation are considered. Mechanisms of interrelation of optimization and simulation models are investigated, and also the algorithm of simulation optimization model development of warehouse system for support of planning processes is described in detail.

In this paper, we discuss the procedure for finding nominal trajectories of the planar five-link bipedal robot with point contact. To this end we use a virtual constraints method that transforms robot’s dynamics to a lowdimensional zero manifold; we also use a nonlinear optimization algorithms to find virtual constraints parameters that minimize robot’s cost of transportation. We analyzed the effect of the degree of Bezier polynomials that approximate the virtual constraints and continuity of the torques on the cost of transportation. Based on numerical results we found that it is sufficient to consider polynomials with degrees between five and six, as further increase in the degree of polynomial results in increased computation time while it does not guarantee reduction of the cost of transportation. Moreover, it was shown that introduction of torque continuity constraints does not lead to significant increase of the objective function and makes the gait more implementable on a real robot.

We propose a two step procedure for finding minimum of the considered optimization problem with objective function in the form of cost of transportation and with high number of constraints. During the first step we solve a feasibility problem: remove cost function (set it to zero) and search for feasible solution in the parameter space. During the second step we introduce the objective function and use the solution found in the first step as initial guess. For the first step we put forward an algorithm for finding initial guess that considerably reduced optimization time of the first step (down to 3–4 seconds) compared to random initialization. Comparison of the objective function of the solutions found during the first and second steps showed that on average during the second step objective function was reduced twofold, even though overall computation time increased significantly.

A previously developed model that describes the dynamics of social tension in a society divided into two groups: the elite and the people was considered. This model took into account the impact of economic situation changes and the elite–people interaction. The model has been modified by including in the equation describing the tension of the people, a term that takes into account the adaptation of the people to the current situation.

The model coefficients estimation is an important task, the solution of which allows obtaining information about the nature of the interaction between elite and people. We believe that the solution of the system of model equations with optimal coefficients is closest to the values of the indicator characterizing social tension. We used the normalized level of homicide rate as an indicator of social tension.

The model contains seven coefficients. Two coefficients characterizing the influence of economic situation changes on elite and people are taken equal to each other and the same for all countries. We obtained their estimations using a simplified model that takes into account only the change in the economic situation and allows an analytical solution.

The Bayesian approach was used to estimate the remaining five coefficients of model for post-Soviet countries. The prior probability densities of the four coefficients for all countries under consideration were taken to be the same. The prior probability density of fifth coefficient was considered to depend on the regime of government (authoritarian or «transitional»). We assumed that the calculated tension matches with the corresponding indicator of tension in cases where the difference between them does not exceed 5%.

The calculations showed that for the post-Soviet countries, a good coincidence was obtained between the calculated values of the people tension and the normalized level of homicide rate. The coincidence is satisfactory only on average.

The following main results was obtained at the work: under the influence of some «significant» events in 40% of post-Soviet countries, there was a rapid change in the nature of interaction between the elite and the people; regional feature have some influence on the elite–people interaction; the type of government does not significantly affect the elite–people interaction; the method for assessing the stability of the country by the value of the model coefficients is proposed.

Based on the time-discrete model, we study the effect of selective proportional harvesting on the population dynamics with age and sex structure. When constructing the model, we assume that the population birth rate depends on the ratio of the sexes and the number of formed pairs. The regulation of population growth is carried out by limiting the juvenile’s survival when the survival of immature individuals decreases with an increase in the numbers of sex and age classes. We consider cases where the harvest is carried out only from a younger age class or from a group of mature females or males. We find that the harvesting of males or females at the optimal level is responsible for changing the ratio of females to males (taking into account the average size of the harem). We show that the maximum number of harvested males is achieved either at such a harvest rate when their excess number is withdrawn and the balance of sexes is established or at such an optimal catch quota at which the sex ratio is shifted towards breeding females. Optimal female harvesting, in which the highest number of them are taken, either maintains a preexisting shortage of adult males or leads to an excess of males or the fixing of a sex balance. We find that, depending on the population parameters for all considered harvesting strategies, the hydra effect can observe, i. e., the equilibrium size of the exploited sex and age-specific group (after reproduction) can increase with the growth of harvesting intensity. The selective harvesting, due to which the hydra effect occurs, simultaneously leads to an increase remaining population size and the number of harvested individuals. At the same time, the size of the exploited group after reproduction can become even more than without exploitation. Equilibrium harvesting with the optimal harvest rate that maximizes yield leads to a population size decrease. The effect of hydra is at lower values of the catch quota than the optimal harvest rate. At the same time, the consequence of the hydra effect may be a higher abundance of the age-sex group under optimal exploitation compared to the level observed in the absence of harvesting.

Although the era of exponential performance growth in computer chips has ended, processor core numbers have reached 16 or more even in general-purpose desktop CPUs. As DRAM throughput is unable to keep pace with this computing power growth, CPU designers need to find ways of lowering memory traffic per instruction. The straightforward way to do this is to reduce the miss rate of the last-level cache. Assuming “non-inclusive cache, inclusive directory” (NCID) scheme already implemented, three ways of reducing the cache miss rate further were studied.

The first is to achieve more uniform usage of cache banks and sets by employing hash-based interleaving and indexing. In the experiments in SPEC CPU2017 refrate tests, even the simplest XOR-based hash functions demonstrated a performance increase of 3.2%, 9.1%, and 8.2% for CPU configurations with 16, 32, and 64 cores and last-level cache banks, comparable to the results of more complex matrix-, division- and CRC-based functions.

The second optimisation is aimed at reducing replication at different cache levels by means of automatically switching to the exclusive scheme when it appears optimal. A known scheme of this type, FLEXclusion, was modified for use in NCID caches and showed an average performance gain of 3.8%, 5.4 %, and 7.9% for 16-, 32-, and 64-core configurations.

The third optimisation is to increase the effective cache capacity using compression. The compression rate of the inexpensive and fast BDI*-HL (Base-Delta-Immediate Modified, Half-Line) algorithm, designed for NCID, was measured, and the respective increase in cache capacity yielded roughly 1% of the average performance increase.

All three optimisations can be combined and demonstrated a performance gain of 7.7%, 16% and 19% for CPU configurations with 16, 32, and 64 cores and banks, respectively.

This paper demonstrates the advantages of using artificial intelligence algorithms for the design of experiment theory, which makes possible to improve the accuracy of parameter identification for an elastostatic robot model. Design of experiment for a robot consists of the optimal configuration-external force pairs for the identification algorithms and can be described by several main stages. At the first stage, an elastostatic model of the robot is created, taking into account all possible mechanical compliances. The second stage selects the objective function, which can be represented by both classical optimality criteria and criteria defined by the desired application of the robot. At the third stage the optimal measurement configurations are found using numerical optimization. The fourth stage measures the position of the robot body in the obtained configurations under the influence of an external force. At the last, fifth stage, the elastostatic parameters of the manipulator are identified based on the measured data.

The objective function required to finding the optimal configurations for industrial robot calibration is constrained by mechanical limits both on the part of the possible angles of rotation of the robot’s joints and on the part of the possible applied forces. The solution of this multidimensional and constrained problem is not simple, therefore it is proposed to use approaches based on artificial intelligence. To find the minimum of the objective function, the following methods, also sometimes called heuristics, were used: genetic algorithms, particle swarm optimization, simulated annealing algorithm, etc. The obtained results were analyzed in terms of the time required to obtain the configurations, the optimal value, as well as the final accuracy after applying the calibration. The comparison showed the advantages of the considered optimization techniques based on artificial intelligence over the classical methods of finding the optimal value. The results of this work allow us to reduce the time spent on calibration and increase the positioning accuracy of the robot’s end-effector after calibration for contact operations with high loads, such as machining and incremental forming.

For the three-axis stabilization of the spacecraft in the orbital coordinate system, including in the indirect equilibrium position, an electrodynamic control method is used based on the simultaneous use of two control torques that affect the dynamics of the spacecraft’s rotational motion in the Earth’s magnetic field (EMF), namely, the Lorentz torque and the torque of magnetic interaction. It is assumed that the spacecraft, equipped with an electric charge with a controlled vector of static moment of charge of the first order and a controlled intrinsic magnetic moment, moves in a Keplerian circular Earth orbit of arbitrary inclination. It was previously shown that combining two control systems, magnetic and Lorentz control, into a single electrodynamic control system (EDCS) makes it possible to successfully solve various problems of controlling the angular motion of spacecraft. Unlike many well-known studies performed for one or another approximate EMF model, this work does not impose restrictions on the accuracy of the EMF approximation. Previous studies have shown the limited capabilities of the EDCS for spacecraft moving in orbits close to the polar ones, due to the presence in this case of such points on the spacecraft trajectory in which it is possible for the lines of action of the geomagnetic induction vector and the spacecraft velocity vector relative to the EMF. Therefore, in this paper, the problem of overcoming these difficulties is posed and solved. A modification of the EDCS is proposed, based, firstly, on optimizing the control of the angular motion of the spacecraft and, secondly, on limiting the maximum value of the modulus of the vector of the center of charge relative to the center of mass of the spacecraft, which must be created during control. A method for selecting parameters for a modified EMF is recommended. The presented results of numerical experiments for spacecraft located in polar and circumpolar orbits not only demonstrate the operability of the proposed modification of the EDCS, but also indicate the possibility of technical implementation of the modified electrodynamic method of three-axis spacecraft stabilization.

New Asynchronous Differential Evolution algorithm is used to determine the parameters of microscopic optical potential of elastic pion scattering on ²⁸Si, ⁵⁸Ni and ²⁰⁸Pb nuclei at energy 130, 162 and 180 MeV.

This article is dedicated to a comparison analysis of optimization methods, in order to perform an interval estimation of electrical energy technical losses in distribution networks of voltage 6–20 kV. The issue of interval evaluation is represented as a multi-dimensional conditional minimization/maximization problem with implicit target function. A number of numerical optimization methods of first and zero orders is observed, with the aim of determining the most suitable for the problem of interest. The desired algorithm is BOBYQA, in which the target function is replaced with its quadratic approximation in some trusted region.