This article proposes an approach to partially use the values of synapses for mutual synchronization of neural networks. Tree parity machines are used as mutually synchronizing neural networks. Confirmation of the completion of mutual synchronization is performed by comparing the hash calculated based on the values of the synapses of neural networks. Partial use of synapse values is proposed to save part of the data from transmission over the network in direct or encrypted form. The study experimentally confirmed the possibility of such an approach and the allowable amount of unused data. Improvements to this method have also been proposed and experimentally tested. The tasks of further research are determined.

A comparative analysis of the existing face recognition systems was conducted; their advantages and disadvantages were revealed. The main functional requirements to the experimental sample of the face recognition software system with a built-in improved model of artificial intelligence are determined, and the use cases are formed. The interface of the experimental system, the main capabilities of the created sample are presented. The results of the experiments are given.

The object of research is the processes of 3D face modeling using the 3D Morphable Model approach. The subject of the work is connected with the improvement of the method of principal components (PCA, principal component analysis) in modeling images of a human face. The purpose of the work is to develop, substantiate and programmatically implement a system for creating a three-dimensional face model from two-dimensional images (photographs). The scientific novelty of the work is to obtain an improved PCA algorithm (principal component analysis). The practical value of the work lies in the universality of the developed algorithm and software implementation, which will allow modeling different faces of people with better quality.

The principles of the face recognition technology building, which are based on the architecture of convolutional neural networks, are considered. Attention is paid to such stages of recognition technology as detection, feature extraction, identification and verification. Also, neural network learning methods, cost functions are reviewed. FaceNet and OpenFace architectures were chosen for further analysis. The OpenFace architecture has been modified through the classifier learning method. TensorFlow and Keras libraries, LFW and Pin Faces network learning datasets were used to build the neural network structure. The ORL Faces dataset was used for testing. The accuracy of training reached 98%. The improvement after modification on the test dataset reached 17%. It was decided to use a modified architecture for implementation in the face recognition system under development.

The most common approaches to face recognition are considered, the latest developments in this area are analyzed, an experiment is carried out on a face recognition system on the Neural Compute Stick 2, the conditions for its necessity are substantiated. The system also was tested on CPU and GPU, as well as system models separately to determine the conditions for increasing performance. The main result is the determination of the need for Neural Compute Stick 2 on personal computers, as well as the optimal conditions for the performance of the face recognition system. The need to use the Neural Compute Stick 2 only for low-performance devices is revealed. The tasks of further research are defined.

This article presents experimental results of evaluating factors affecting synchronization time of tree parity machines. Tree parity machines are proposed as a modification of the symmetric encryption algorithm. One of the advantages of the method consists in using the phenomenon of mutual synchronization of neural networks to generate an identical encryption key for users without the need to transfer it. As a result, the factors influencing the synchronization time of neural networks and the level of key cryptographic strength were determined. The degree of influence factors was found out experimentally. The influence of the learning rule on timing and stability of synchronization of neural networks was also determined. As a result, it was determined that the best rule for mutual learning of neural networks is Hebb’s rule, and when the architecture of neural networks becomes more complex, the number of hidden neurons should be increased first. The tasks of further research are defined.

The most common methods for reconstruction of 3D face models are considered, their quantitative estimates are analyzed and determined, the most promising approach - 3D Morphable Model is highlighted. The necessity of its modification is substantiated in order to improve the results of reconstruction based on the analysis of the main components and the use of a neural network. One of the advantages of using the 3D Morphable Model with the analysis of the main components is the representation of only a probable solution, when the space of solutions is limited, thereby simplifying the problem being solved. While the original approach involves manual initialization. The main result, is an approach for creating three-dimensional face models from their 2D images, having least time and satisfactory root mean square error. The tasks of further research are determined.

The need for real time image generation of landscapes arises in various fields as part of tasks solved by virtual and augmented reality systems, as well as geographic information systems. Such systems provide opportunities for collecting, storing, analyzing and graphically visualizing geographic data. Algorithmic and hardware software tools for increasing the realism and efficiency of the environment visualization in 3D visualization systems are proposed. This paper discusses a modified path tracing algorithm with a two-level hierarchy of bounding volumes and finding intersections with Axis-Aligned Bounding Box. The proposed algorithm eliminates the branching and hence makes the algorithm more suitable to be implemented on the multi-threaded CPU and GPU. A modified ROAM algorithm is used to solve the qualitative visualization of reliefs’ problems and landscapes. The algorithm is implemented on parallel systems—cluster and Compute Unified Device Architecture-networks. Results show that the implementation on MPI clusters is more efficient than Graphics Processing Unit/Graphics Processing Clusters and allows real-time synthesis. The organization and algorithms of the parallel GPU system for the 3D pseudo stereo image/video synthesis are proposed. With realizing possibility analysis on a parallel GPU-architecture of each stage, 3D pseudo stereo synthesis is performed. An experimental prototype of a specialized hardware-software system 3D pseudo stereo imaging and video was developed on the CPU/GPU. The experimental results show that the proposed adaptation of 3D pseudo stereo imaging to the architecture of GPU-systems is efficient. Also it accelerates the computational procedures of 3D pseudo-stereo synthesis for the anaglyph and anamorphic formats of the 3D stereo frame without performing optimization procedures. The acceleration is on average 11 and 54 times for test GPUs.