Self-learning Machine Research Articles

Self-Learning Machines Based on Hamiltonian Echo Backpropagation

Self-Learning Machines Based on Hamiltonian Echo Backpropagation

Safe contextual Bayesian optimization integrated in industrial control for self-learning machines

Intelligent manufacturing applications and agent-based implementations are scientifically investigated due to the enormous potential of industrial process optimization. The most widespread data-driven approach is the use of experimental history under test conditions for training, followed by execution of the trained model. Since factors, such as tool wear, affect the process, the experimental history has to be compiled extensively. In addition, individual machine noise implies that the models are not easily transferable to other (theoretically identical) machines. In contrast, a continual learning system should have the capacity to adapt (slightly) to a changing environment, e.g., another machine under different working conditions. Since this adaptation can potentially have a negative impact on process quality, especially in industry, safe optimization methods are required. In this article, we present a significant step towards self-optimizing machines in industry, by introducing a novel method for efficient safe contextual optimization and continuously trading-off between exploration and exploitation. Furthermore, an appropriate data discard strategy and local approximation techniques enable continual optimization. The approach is implemented as generic software module for an industrial edge control device. We apply this module to a steel straightening machine as an example, enabling it to adapt safely to changing environments.

Capture Based Trust Dependence framework for authorized node identification in mobile agent systems

A mobile agent is a self-learning machine entity that uses the system infrastructure to keep running in another remote zone, check and compile the results, interact with various locations and return to his home site after completing the relegated activities. Mobile Agent-based solutions for the testing community have grown in popularity and are now used in a variety of fields, including boardroom management, electronic commerce, renewable energy and power management. Addition to these applications, Broadband Interactive Sensors, network performance improvement, disseminated knowledge mining, multimedia, human monitoring, surveillance, affective computing, weather and environment, e-learning and semantic web administrations are only a few of the topics covered. In an extremely non trusty environment, focus should be taken to shield the portable operator from acquiring altered. Existing works on mobile agent frameworks with very surprising instruments does not offer complete security. In this paper a capture based trust dependence framework is proposed for identification of authorized or trust nodes inside mobile agents systems. Here we consider the mobile adhoc networks for computational analysis of network performance using network simulator. The framework provides efficient results in identification of authorized nodes.

Различия и сходства понятий «промышленные кластеры» и «промышленные экосистемы»

The article considers the modern realities of the current period of time caused by the fourth industrial revolution, the technology of which is inextricably linked to the digitalization. The existing trends can lead to a decrease in social well-being due to the loss of jobs due to the introduction of self-learning machines in production, as well as environmental degradation due to the increased demand for natural resources. These possible consequences actualize the search for models for the organization of production, among which in foreign practice are recognized industrial clusters and ecosystems. In the scientific environment the definition of criteria allowing to differentiate these concepts is an insufficiently studied question. The similarities and differences between the definitions of industrial clusters and industrial ecosystems have been analyzed. There is considered the ecosystem approach for industry in terms of Industry 4.0 and digitalization. Using the method of content analysis the publications were distributed in the system of National Electronic Library (eLibrary) from 2004 to 2020. There were defined the key criteria for a comparative analysis of the studied concepts: dynamism, flexibility, openness to outside challenges, objectives, participants, territory, process chain, innovation, organizational structure. The relationship between goals and effects for industrial clusters and industrial ecosystems has been illustrated. It has been inferred that there are significant differences in the conceptual apparatus of an industrial ecosystem and an industrial cluster. It has been substantiated that industrial ecosystems are an evolutionarily more perfect form of organizing the system of interaction between industrial organizations than industrial clusters

Artificial intelligence and machine learning for internet of things

For decades, humans have been intrigued by the concept of an intelligent and independent self-learning machine. The idea behind Machine Learning (ML) is to simplify the development of analytical models such that, with the help of available data, algorithms can learn continuously. Internet of Things (IoT) enabled devices are the major sources of data generation with a number of multiple modalities and differing data consistency, defined by velocity in terms of time and position dependence. Intelligent processing and analysis of this generated data (Big Data) is the key to developing smart IoT applications. ML may be used in cases where the desired effect is defined (supervised learning) or where data itself is not defined beforehand (unsupervised learning) or where learning is the outcome of the interaction among the learning model and the environment (reinforcing learning). In this chapter, we present and discuss a taxonomy of machine learning algorithms that can be used with IoT. Furthermore, how different machine learning techniques are used to derive higher-level information from the data is illustrated. Lastly, we investigate, what are the real-world IoT data characteristics that involve an interpretation of the data?

Two Support Vector Machine Methods for Image Noise Filter

To solve the fatal limitation of SVMs based on the pre-known, two new image noise filter methods - Support Vector Machine based on self-learning machine (SLM-SVM) were presented, which is working better than many other non-linear filters, such as median filters and adaptive filters. A series of comparison and working parameters would be explained, performance of test showed these methods can dealing with the noise in a totally unknown image. And filter out more than 99% noise pixels from an image. In the same time, this kind of filter is still immature due to its long runtime (more than 2 second) and some error cases during testing, some recommendation and prediction were proposed.

Artificial Intelligence, the Need of the Hour

Fourth Industrial revolution is currently sweeping the high-income countries (HIC) with Artificial Intelli- gence (AI) based automation affecting virtually every aspect of life. The term AI was first coined by McCar- thy in 1956. It was not until 2000s that AI began to thrive. The evolution of AI into the current status occurred in the last decade owing to the enhanced computing power using Graphic Processing Units (GPU), development of high-powered computer languages, and the emergence of the Big Data. The latter is generated through wireless communication between ‘Smart’ sensors/devices and self-learning machines. The word ‘smart’ is applied to any device that has memory and is able to connect with data networks such as the internet and the processors. In the last few years, there has been exponential growth in AI applications. This can be judged by the projec- tion that the AI field will add $ 15 Trillion to global economy, by the year 2030, up from $ 600 Million in 2016. This will occur mostly in the HIC. The adoption of AI by low- and middle-income countries (LMIC) lags far behind that of HICs. The LMICs would miss out in the economic benefits, further widening the global inequalities. Machine Learning and Deep Learning are branches of AI that are beginning to form the basis of the automation of financial and business decisions, and are the tools of self-driving cars, industrial produc- tion, data analytics, quality improvement and health- care processes to name a few. In healthcare, some of the AI applications have shown to enhance patient care, reduce medical errors, support clinical and administrative decision making, automate equipment maintenance and help reduce operational cost. For instance, AI led cost reductions achieved up to 25 percent drop in the length of hospital stay and up to 91 per cent reduction in admissions to step down facili- ties. In the United States alone, by the year 2026, AI in healthcare is estimated to realize $150 billion in annual cost savings.

Methodological foundations of the fourth industrial revolution research

The article conceptualises the stages that humanity has passed on the way of technological development: the transition from agricultural production to the fi rst steam engines, from mastering the power of steam to electricity, from electricity to digital technologies; and, fi nally, the transition that we are witnessing: from digital technologies to the creation of complex interconnected cyber-physical systems based on autonomous self-learning machines.

On Fiduciary Relationship with Artificial Intelligence Systems

This article aims to address the theoretical and practical issues of fiduciary relationship with artificial intelligence systems. The findings point out that human can establish conditional fiduciary relationships with autonomous artificial intelligence systems by unilateral recognition, or authorization by consent. Artificial intelligence systems can have a social personality when authorized by humans. Although AI cannot sue and be sued without recognition of a legal personality. This paper provides an original contribution to the existing knowledge on the nature of self-learning machines and international regulation of autonomous artificial intelligence systems.

A Review on Importance of Active Learning Neural Network for Grid Connected Renewable Energy System

As Renewable Energy Sources are being used everywhere to minimize the burden of power plants and global emission. Therefore, Integrating Renewable energy with the grid is the trend nowadays in the Power distribution system. Power network get congested due to change in power flow from unidirectional to bidirectional. This paper discusses the importance of self-learning machine, in mitigating the power quality issues while integrating renewable energy with the power grid. Renewables energy are randomness in nature, due to its recurrent characteristic power network get effected. Different approaches like advanced forecasting, load flow analysis etc. are implemented using Artificial Intelligence (AI) techniques to improve the power network. AI machine can perform only those tasks which are pre-defined in its internal memory. Huge amount of data and time are required to train an AI machine. Whereas Active Learning Neural Network (ALNN) uses a self-learning approach to train itself through some model data. This paper mainly focusses on the importance of using ALNN in mitigating the power quality issues much better.

Algorithmic Copyright Enforcement and AI: Issues and Potential Solutions through the Lens of Text and Data Mining

Although digitalization and the emergence of the Internet has caused a long-term crisis for copyright law, technology itself also seems to offer a seemingly ideal solution to the challenges of digital age: copyright has been a major use case for algorithmic enforcement from the early days of digital rights management technologies to the more advanced content recognition algorithms. These technologies identify and filter possibly infringing content automatically, effectively and often in a preventive fashion. These methods have been criticized for their shortcomings, such as the lack of transparency, bias and the possible impairment of fundamental rights. Self-learning machines and semi-autonomous AI have the potential to offer even more sophisticated and expeditious enforcement by code, however, they could also aggravate the aforementioned issues. As the EU legislator envisions to make the use of such technologies essentially obligatory for certain online content sharing service providers (via the infamous Article 17 of the directive on copyright in the digital single market), the assessment of the situation in light of future technological development has become a current topic.This paper aims to identify the main issues and potential long-term consequences of creating legislation that practically requires the employment of such filtering algorithms as well as their solutions. This paper focuses on the potential role a broad copyright exception for text and data mining could play in counterbalancing the issues associated with algorithmic enforcement.

Integrating Product-Service Systems into the manufacturing industry: Industry 4.0 perspectives

Industry 4.0 is the current trend in the technologies used in exchange of data in manufacturing. Product-service system is a concept that provides self-aware and self-learning machines, and consequently improve overall performance and maintenance management. Companies continually strive to increase production, but in recent years, the effects of this effort have demonstrated that provision of products alone with less resilient technology-enabled service industry is insufficient to remain competitive. The importance of leveraging flexibility and capabilities offered by cloud computing is inevitable, but adapting prognostics and health management algorithms to efficiently implement current data management technologies was research and develop in this study. In this paper, we present an integrated product-service system as a potential concept from a life-cycle perspective to reduce disruption, costs and associated services such as autonomous maintenance, through exchange of data to understand equipment degradation using sensors application, to aid timely services intervention for manufacturing processes transformation. The results is the knowledge base servitude backed by machine clusters generated data for different working conditions to improve operational efficiency using Cloud IoT to connect, process, store, and analyse data on the dashboard, locally and in the cloud for future health assessment.

Proto-RNA, The First Self-Learning Machine

Proto-RNA, The First Self-Learning Machine

Development of the self-learning machine for creating models of microprocessor of single-phase earth fault protection devices in networks with isolated neutral voltage above 1000 V

The paper proposes the development of a self-learning machine for creating models of microprocessor-based single-phase ground fault protection devices in networks with an isolated neutral voltage higher than 1000 V. Development of a self-learning machine for creating models of microprocessor-based single-phase earth fault protection devices in networks with an isolated neutral voltage higher than 1000 V. allows to effectively implement mathematical models of automatic change of protection settings. Single-phase earth fault protection devices.

Securities market automation from standards to self-learning machines: Current state and future perspectives

Over the last two decades, key aspects of the financial industry have been automated to a substantial degree. While most progress in automation has come from traditional technologies, recent advances in machine learning, artificial intelligence and robotics are likely to accelerate the pace. In such a context, Distributed Ledger Technologies, Robo-Advisors and cognitive tools are creating a foundation for solving major problems faced by the industry. This paper provides an overview of the capabilities and limitations of these technologies and the challenges that await market participants who want to embrace and implement them. It draws attention to the importance of collaboration, governance, standards and market practice harmonisation in order to successfully deploy these technologies in a multi-party, globalised network environment.

Raman Spectroscopy-Compatible Inactivation Method for Pathogenic Endospores

Micro-Raman spectroscopy is a fast and sensitive tool for the detection, classification, and identification of biological organisms. The vibrational spectrum inherently serves as a fingerprint of the biochemical composition of each bacterium and thus makes identification at the species level, or even the subspecies level, possible. Therefore, microorganisms in areas susceptible to bacterial contamination, e.g., clinical environments or food-processing technology, can be sensed. Within the scope of point-of-care-testing also, detection of intentionally released biosafety level 3 (BSL-3) agents, such as Bacillus anthracis endospores, or their products is attainable. However, no Raman spectroscopy-compatible inactivation method for the notoriously resistant Bacillus endospores has been elaborated so far. In this work we present an inactivation protocol for endospores that permits, on the one hand, sufficient microbial inactivation and, on the other hand, the recording of Raman spectroscopic signatures of single endospores, making species-specific identification by means of highly sophisticated chemometrical methods possible. Several physical and chemical inactivation methods were assessed, and eventually treatment with 20% formaldehyde proved to be superior to the other methods in terms of sporicidal capacity and information conservation in the Raman spectra. The latter fact has been verified by successfully using self-learning machines (such as support vector machines or artificial neural networks) to identify inactivated B. anthracis-related endospores with adequate accuracies within the range of the limited model database employed.

A new Hierarchical Pattern Recognition method using Mirroring Neural Networks

this paper, we develop a hierarchical classifier (an inverted tree-like structure) consisting of an organized set of each of which is actually a module that performs a feature extraction and an associated classification. We build each of such blocks by coupling a Mirroring Neural Network (MNN) with a clustering (algorithm) wherein the functions of the MNN are automatic data reduction and feature extraction which precedes an unsupervised classification. We then device an algorithm which we name as a Tandem Algorithm for the self-supervised learning of the MNN and an ensuing process of unsupervised pattern classification so that an ensemble of samples presented to the hierarchical classifier is classified and then sub-classified automatically. This tandem process is a two step process (feature extraction/data reduction and classification), implemented at each block (module) and can be extended level by level in the hierarchical architecture. The proposed procedure is practically demonstrated using 2 example cases where in a collage of images consisting of faces, flowers and furniture are classified and sub classified automatically. We expect that this kind of architecture will be very useful in the development of efficient and powerful self-learning machines in the future.