Vol 15, No 2 (2024)
Distributed ALUs based on photonic switches
Abstract
The article examines a photonic network with distributed control, consisting of several nodes connected by a common channel, in which, during the transmission of one number, a single operation is performed on numbers that are transmitted in parallel by all nodes. Such operations as summing or finding the maximum (minimum) of numbers transmitted sequentially across binary bits are considered. It is assumed that the bits of numbers are transmitted by paraphase optical signals, and the common channel is built from photonic switches of these signals.
Program Systems: Theory and Applications. 2024;15(2):3-19
3-19
Gesture control of small unmanned aerial vehicle flight
Abstract
The problem of constructing gesture commands for controlling a small unmanned aerial vehicle, such as a quadcopter, is considered. Commands coming from a video camera are identified by a classifier based on a convolutional neural network, and the multimodal control interface equipped with an intelligent solver converts them into control commands for the quadcopter. Neural networks from the Ultralytics neural network library allow selecting targets in a frame in real-time. The commands are sent to a specialized program on a smartphone, developed on the basis of DJI SDK flight simulators, which then sends commands via the remote control channel.The quality of recognition of developed gesture commands for DJI Phantom 3 standard edition quadcopters is investigated, and a brief guide in the form of operator work scenarios with unmanned vehicles is provided. The prospects of gesture control of several vehicles in extreme conditions have been revealed, considering the complex safety challenges of joint flight and interaction of aircraft in confined space.
Program Systems: Theory and Applications. 2024;15(2):21-36
21-36
A systematic review of methods for deriving metamorphic relations
Abstract
Metamorphic testing is one of the most effective methods of testing programs with the test oracle problem. This problem declares that it is impossible to know whether the test answer is correct for one reason or another. Metamorphic testing uses metamorphic relations to check the program correctness. Metamorphic relation is a function of several test inputs and corresponding outputs of the program. Developing metamorphic relations can be a non-trivial task.This systematic review is dedicated to identifying general derivation techniques for metamorphic relation as well as techniques pertinent to particular domains. As a result, we propose a classification of techniques into six main types and compile a comparative table of input data transformations for testing tasks in different domains. Findings of this review will help researchers to apply metamorphic testing in practice.
Program Systems: Theory and Applications. 2024;15(2):37-86
37-86
The use of distributed computing in domain modeling in universal syllogistics
Abstract
We consider the algorithmic aspects of calculations in the logical-semantic models of universal syllogistics $L_{S_{2}}$. This non-classical propositional logic is based on an algebraic system containing a Boolean algebra of sets and two relations between sets: $\subset$ and $=$. Its closest analogue is Aristotle's syllogistics, the model of which is an algebraic system with a Boolean algebra of sets and one relation $\subset$. The disadvantage of syllogistics based on an algebraic system with a single relation $\subset$ is the ambiguity of the interpretation of their formulas and atomic judgments.In this work, by a logical-semantic model of a subject area we understand the totality of the universal syllogistic formula $L_{S_{2}}$ and its semantic meaning, which is a finite set of non-negative integers. We propose an algorithm for computing semantic value of a conjunctive well-constructed formula $L_{S_{2}}$, which has a high level of parallelism at the task level, at the data level, and at the level of algorithms realizing operations on constituent sets. Due to the peculiarities of union, intersection and complement operations over finite sets, all the processes of their computation and solution of subtasks occur at the bit level and, as a rule, are efficiently implemented in algorithmic languages. In the proposed algorithm, the transition to the bit level and back is realized by a set of software tools.
Program Systems: Theory and Applications. 2024;15(2):87-112
87-112
Application of neural networks of Siamese architecture in problems of classifying products of various categories on supermarket shelves
Abstract
This paper presents a study on the application of Siamese architecture neural networks in problems of classifying various food products on the shelves of department stores. Siamese networks are a special class of neural network architectures that combine two convolutional subnets. This type of neural networks is often used in object matching problems and has an important advantage over traditional convolutional neural networks, namely the absence of the need for a large amount of training data. During the work, we generated our own data set, including five different product categories. As a result, it was possible to achieve a tonality of 97.5% during training.
Program Systems: Theory and Applications. 2024;15(2):113-137
113-137
New generation of GPGPU and related hardware: computing systems microarchitecture and performance from servers to supercomputers
Abstract
An overview of the current state of GPGPUs is given, with orientation towards their using to traditional HPC tasks (and less to AI). The basic GPGPUs in the review include Nvidia V100 and A100. Nvidia H100, AMD MI100 and MI200, Intel Ponte Vecchio (Data Center GPU Max), as well as BR100 from Biren Technology are considered as new generation GPGPUs. The important for HPC and AI tasks microarchitecture and hardware features of these GPGPUs, as well as the most important additional hardware for building computer systems with GPGPUs, that are CPUs specialized (albeit only possible for the initial period of their use) for working with the new generation of GPGPUs and interconnects — are analyzed and compared. Brief information is given about the servers (including multi-GPUs) using them, and new supercomputers (using these GPGPUs), where data on the achieved performance when working with GPGPUs was obtained.The SDK of GPGPU manufacturers and software (including mathematical libraries) from other firms are briefly reviewed. Examples are given that demonstrate the tools of widely used programming models that are important for achieving maximum performance, while contributing to the non-portability of program codes to other GPGPU models.Particular attention is paid to the possibilities of using tensor cores and their analogues in modern GPGPUs from other companies, including the possibility of using calculations with reduced (relative to the standard for HPC FP64 format) and mixed precision, which are relevant due to the sharp increase of the achieved performance when using them in GPGPU tensor cores. Data is analyzed on their “real-world” performance in benchmarks and applications for HPC and AI. The use of modern batch linear algebra libraries in GPGPU, including for HPC applications, is also briefly discussed.
Program Systems: Theory and Applications. 2024;15(2):139-473
139-473
Architecture of interaction in the digital medical ecosystem
Abstract
In the field of medical informatics, there is a steady trend towards the formation of a complex multicomponent ecosystem. The problems of interaction and integration of ecosystem components come to the forefront: medical, laboratory, radiological systems, EGISZ, EMIAS, MDLP, various registers and services, including those implementing AI approaches to data processing and problem solving. Patients are in dire need of personal offices that integrate their medical data, patients become active participants in the ecosystem. Integration tasks have to be solved in a highly heterogeneous information environment, when it becomes unattainable to ensure synchronous interactive interaction between ecosystem participants. For individual applications, a flexible combination of both synchronous and asynchronous interaction is required, selected situationally based on specific time delays and interaction characteristics.The article proposes a special architecture that allows for synchronous and asynchronous interaction between ecosystem participants. Adapting software designed only for synchronous interaction to an asynchronous architecture does not require a radical redesign of the software. The approach was worked out using the example of adapting the MDLP MIS Interin module to work in the internal secure network of the multidisciplinary medical center of the Bank of Russia. The proposed architecture can be used by software developers in other fields of activity, where there is an active development of ecosystems, accompanied by an increase in integration interactions. Key Words and Phrases:.
Program Systems: Theory and Applications. 2024;15(2):475-492
475-492