Harvester open publications of NAS Ukraine
Harvester open publications of NAS UkraineNot a member yet
2350 research outputs found
Sort by
Some aspects of software engineering for AI-based systems
No full textAI-based software systems are rapidly spreading in various business areas. In this context, the unavoidable convergence of the Software Engineering and Artificial Intelligence and Machine Learning (AI/ML) disciplines is considered an obvious and one of the following significant challenges within the engineering process. The life cycle, models, and technologies of AI/ML elements are pretty specific, and this should be considered in software engineering to ensure their performance and compliance with business needs. AI/ML applications have some distinct characteristics compared to traditional software applications. Thus, several challenges and risk factors regarding AI/ML applications appear to software developers. To study the common challenges in AI/ML application development, we used two different perspectives: software engineering and machine learning. AI/ML applications, like other software systems, need a well-defined software engineering process for their development and maintenance. We discussed challenges and recommendations for different phases of the software development life cycle for ML applications, particularly requirement engineering, design, implementation, integration, testing, and deployment. AI/ML application development has specific aspects to consider as a software development project. We discussed the characteristics and recommendations concerning problem formulation, data acquisition, preprocessing, feature extraction, model building, evaluation, model integration and deployment, model management, and ethics in AI/ML development. In the work, there were formulated recommendations for each analyzed challenge that should be useful for software developers. The next stage of this research is the compilation of detailed systematic guidelines for the software development process for AI/ML systems.Prombles in programming 2022; 3-4: 99-106
Ontological system processing of databases of scientific publications
Get PDFDevelopment of theories, methods and algorithms for the discovery and formation of new knowledge always was one of the most important tasks for any researcher, especially if they actively working on creation of new scientific publications. There is no universal language to describe formally concepts (knowledge) and systemology of transdisciplinary scientific research. Because of this, researchers have a set of urgent problems, and one of them is the way of speeding up the process of finding information (in the form of cognitive-structure) in their own sources. Ontological system for processing of databases of scientific publications created to solve this problem for a researcher, who have from tens to hundreds of scientific papers published. We are unaware of search systems, which would provide the same information for a researcher in such a short time. Ontological system implements technologies of Information Retrieval and Knowledge Discovery in Databases with accent on technologies and instruments such as Semantic Web and cognitive graphics. Development of such ontological system have three stages. On the first stage instruments for system development created, methods and algorithms of interaction between system components "User ¾ Knowledge engineer ¾ Remote endpoint", also data added to the system at this stage. On the second stage task of multimedia presentation for conceptual and figurative structures, described in scientific documents, solved. Gaining new knowl- edge problem solved on the third stage.Prombles in programming 2022; 3-4: 161-17
Applying Description Logics with Concrete Domains to Solve the Problems of Semantic Web Services Discovery and Composition
No full textThis research is one of the branches of a general scientific and applied problem, namely, the problem of using and developing ontological approaches with descriptive logics (DL) apparatus to obtain the effective automated solution of complex business task based on service-oriented architecture. The base idea of the proposed method is the maximal ontologization, that is, all aspects of the task are determined by means of DL ontologies. The proposed ontology system is based on five basic types of ontologies, namely:(1) an ontology of general entities for defining the characteristics of the service’s context, (2) an ontology of the application domain to formalize and unify semantical descriptions of web services, (3) the top-level ontology of web services for the DL formalization of the definition of the web services in general, (4) the ontologies of the web services tasks, which provide a formal representation of the tasks on all stages of web services life cycle, and (5) the ontologies of specific web services. All listed ontologies are related. Such, ontologies of the last type include links to all other above-mentioned ontologies and, together with them, provide a unified formally defined semantic resource, which enables the automatic solution of web service tasks. The focus of this study is the applying descriptive logics with concrete domains to the formalization of such basic semantic web services tasks as the web services discovery and composition. In this article, the ontologies of the web services tasks are defined as an extension of the general top-level web service ontology with special complex concepts. Their elements are pairs of “web service-goal (request)”/”web service-web service”, which are compared according to the task being solved. Also, on their basis, special complex concepts are determined, which specify a set of found similarities between existing services and a goal in the discovery task, or «predecessor-follower» type responsibilities of web services to build a composite service. The proposed TBox statements are constructed using descriptive logic with concrete domains (). For this purpose, a concrete domain is defined as a pair of sets: (1) the union of sets of possible subsets of the set of individuals of the Parameter concept and possible subsets of the set of individuals of the Context concept, and (2) a set of predicates defined on it. Such definition of the concrete domain provides possibility of using only functional roles to construct the DL statements of complex concepts DiscoveryServices and CompositionObject for determining corresponding web services tasks. Note, the defined tasks and web services ontologies cover only simplified model of the representation of web services, in particular, functional Input/Output model extended by contexts descriptions. They provide the basis for further improvement by expanding this model and involving additional characteristics of both functional and process models of representation to the specified task ontologies. This will certainly increase the accuracy of the obtained result, but is the subject of further research.Prombles in programming 2022; 3-4: 207-21
Automated generation of programs for a class of parametric neuroevolution algorithms
No full textThe facilities of algebra of hyperschemes are applied for automated generation of neuroevolution algorithms on an example of a binary multiplexer evaluation problem, which is a part of the SharpNEAT system. SharpNEAT is an open-source framework developed in C# programming language, which implements a genetic neuroevolution algorithm for the .NET platform. Neuroevolution is a form of artificial intelligence, which uses evolution algorithms for creating neural networks, parameters, topology, and rules. Evolution algorithms apply mutation, recombination, and selection mechanisms for finding neural networks with behavior that satisfies to conditions of some formally defined problem. In this paper, we demonstrate the use of algebra of algorithms and hyperschemes for the automated generation of evaluation programs for neuroevolution problems. Hyperscheme is a high-level parameterized specification of an algorithm for solving some class of problems. Setting the values of the hyperscheme parameters and further interpretation of a hyperscheme allows obtaining algorithms adapted to specific conditions of their use. Automated construction of hyperschemes and generation of algorithms based on them is implemented in the developed integrated toolkit for design and synthesis of programs. The design of algorithms is based on Glushkov systems of algorithmic algebra. The schemes are built using a dialogue constructor of syntactically correct programs, which consists in descending design of algorithms by detailing the constructions of algorithmic language. The design is represented as an algorithm tree. Based on algorithm schemes, programs in a target programming language are generated. The results of the experiment consisting in executing the generated binary multiplexer evaluating program on a cloud platform are given.Prombles in programming 2022; 3-4: 301-31
Methods and software for significant indicators determination of the natural language texts author profile
Get PDFMethods for the formation and optimization of author profiles are presented. The author profile is an image – a vector in a multidimensional space, which components are author’s texts measurements by a number of methods based on 4-grams, stemming, recurrence analysis and formal stochastic grammar. The author’s profile is a model of his language, including vocabulary, sentence syntax features. A comparative analysis of each of the methods effectiveness is carried out. By means of the genetic algorithm, a reduced profile of the author is formed. Insignificant indicators are excluded, which allows to reduce their number by 20%. The reduced author’s profile contains attributes that are significant for this author and is an effective attribution of a particular author.Prombles in programming 2023; 3: 22-2
Insertion semantics of quantum interactions
No full textThe rapid development of the chemical industry and science and new challenges in the field of health care put forward increased demands for the development of the theory of organic and inorganic chemistry, biochemistry and biophysics, the search and implementation of new modelling and analysis methods, and the improvement of technological processes.One of the safe and fast methods of researching the properties and behavior of new materials and tools is the modelling of relevant experiments, in particular, computer molecular modelling based on mathematical models.Modelling the interactions between micro and macromolecules at the quantum level allows us to manipulate the substances’ electronic, magnetic, optical and other characteristics and consider the possibilities of creating new chemical bonds, molecular structures, phase transitions, quantum states, and so on.Accordingly, the main idea of our research is to apply the technology of algebraic modelling and quantum-chemical apparatus for the simulation and verification of experiments in physics, chemistry, and biology areas.The use of formal algebraic methods allows proving properties and finding relevant scenarios for the effective analysis of the behavior of various objects in real-time, considering not individual scenarios but sets of possible behaviors.At this research stage, we have developed a methodology for formalization complex organic and inorganic substances, chemical processes and reactions based on the formalization of the interaction of atoms and molecules at the level of quantum interactions.Prombles in programming 2023; 4: 65-7
Artificial intelligence in cloud-based mobile radar computing
Get PDFThe introduction of Artificial Intelligence into mobile radar computing based on cloud resources has made it possible to combine radar resources at the stage of receiving, processing and presenting information. The radar system has become integral and flexible. The convergence of mobile applications in portable devices with cloud computing is a revolution in the organization of distributed computing. Attention is paid to the architecture of the client part as a neurocomputer distributed in space with deep learning capabilities. Providing analysis of radar data in Real Time for a mobile platform is very important and its implementation with packet data transmission at different stages accelerate the analysis process.Problems in programming 2023; 2: 49-5
Novelty search in neuroevolution for end effector positioning
No full textThe article considers the use of the neuroevolution algorithm for neural network policies search when creating a controller for a robotic arm, in particular for the subtask of positioning the end effector. Neuro-evolution is a family of machine learning methods that use evolutionary algorithms by imitating the process of natural selection. This approach has been found to be particularly effective for the positioning task, where the final position can be achieved in many optimal ways and therefore requires reinforcement learning. It is noted that the final result of neuroevolution is an optimal network topology, which makes the model more resource-efficient and easier to analyze. The paper considers the process of neural network policy search for controlling a two-dimensional robot with two links. According to the results of the experiments, an increase in the efficiency of the best solution found using novelty search for the NEAT algorithm is noted compared to the NEAT algorithm without novelty search. It was established that the proposed approach allows to obtain an effective neural network policy, which has a minimal configuration, which will allow to increase the speed of the controller, that is critical for the operation of a real system. Thus, the use of novelty search as a method of optimizing the neuroevolutionary process to solve the positioning problem allows to increase the efficiency of the learning process and obtain the optimal network topology.Prombles in programming 2023; 3: 49-5
Toward software engineering ecosystems definition
Get PDFNowadays, the fundamental science of software engineering is being formed. It should represent knowledge that meets the requirements of the concept of sustainable development. This fundamental science could be named as the Software Engineering Ecology. Along with others, it should include a section containing knowledge about software engineering ecosystems. This section of the future science has been intensively developing for more than fifteen years. However, today, there is no consen- sus among researchers regarding the definitions of the software ecosystem. Naturally, it does not contribute to the creation of an appropriate section of emerging science. All authors offer only a software ecosystem, considering it in different contexts and defining it in different ways. The term “software ecosystem” is now used to refer to a wide range of ecosystems that are actually software engineering ecosystems. The purpose of this paper is to propose a basis for defining software engineering ecosystems. By analogy with the concepts of the landscape and the trophic chain of biological ecosystems, the concepts of software landscape and software engineering value chain are proposed. Based on these concepts, the diversity of software engineering ecosystems is shown. A model of the software engineering ecosystems and a classification of the software engineering ecosystems are proposed.Prombles in programming 2022; 3-4: 92-9
Algebraic approach to the analysis of legal documents
Get PDFIn this study, problems regarding the analysis of law articles and their conformance to legal documents are considered. The algebraic approach is used for the formal verification of legal texts that is presented as specification of behavior algebra. Having a formal presentation of legal texts allows for the application of algebraic methods, such as symbolic modeling, automatic proving of statements and algebraic matching. The approach was implemented for Ukrainian-language legal documents to detect inconsistency, incompleteness, and prove conformance. The analysis of legal texts has been implemented for Tax code and examples of inconsistency of some statements were demonstrated in the paper. Another experiment has been performed for checking of conformance of court statements, agreements, licenses, new laws, taxation acts to the articles of law with usage of methods of algebraic matching. The formalization of legal documents, like court verdicts, shall be implemented fully automatically, and the existing database with such documents shall provide the possibility to use big data technologies and machine learning. In this study, we consider the use of the algebraic approach in the analysis of legal requirements and law artifacts within the scope of the Insertion Modeling System (IMS). The subject of this research is the laws created by the legislature and the artifacts of legal activity, such as lawsuit decisions, agreements, licenses, and juridical cases. The problem to be considered is the analysis of legal documents for conformance with the law and the analysis of the law’s articles, checking for inconsisten- cies, incompleteness, and other interested properties. In this paper, we consider every element of the technology in detail, and we describe the methods and results of the experiments.Prombles in programming 2022; 3-4: 117-12