Self-learning Algorithms Research Articles

Neural Network Algorithms in Current Processes of Transformation of Traditional Worldview and Ideological Systems.

This study is devoted to identifying potential risks, challenges, and threats of intensive development and active implementation of artificial intelligence technologies and in particular self-learning neural network algorithms in the current practice of socio-political communications associated with active information impact on mass consciousness. This study proves that contemporary digital technological transformations and the active spread of digital technologies most directly influence the transformation of traditional worldview systems and ideologies. Based on the results of the study, it is concluded that with the help of digital tools, there is a significant information and communication impact on mass consciousness in terms of influencing traditional worldview and ideological systems. In this regard, the author puts forward the thesis about the need to form and maintain the digital sovereignty of contemporary states as a key conclusion. It is viewed as a critical condition for preserving and protecting national worldview and ideological spaces in the conditions of current global technological transformations

Droplet Shape Representation Using Fourier Series and Autoencoders

The shape of liquid droplets in air plays an important role in the aerodynamic behavior and combustion dynamics of miniaturized propulsion systems such as microsatellites and small drones. Their precise manipulation can yield optimal efficiency in such systems. It is desired to have a minimal representation of droplet shapes using as few parameters as possible to automate shape manipulation using self-learning algorithms, such as reinforcement learning. In this paper, we use a neural compression algorithm to represent, with only two parameters, elliptical and bullet-shaped droplets initially represented with 200 points (400 real numbers) at the droplet boundary. The mapping of many to two points is achieved in two stages. Initially, a Fourier series is formulated to approximate the contour of the droplet. Subsequently, the coefficients of this Fourier series are condensed to lower dimensions utilizing a neural network with a bottleneck architecture. Finally, 5000 synthetically generated droplet shapes were used to train the neural network. With a two-real-number representation, the recovered droplet shapes had excellent overlap with the original ones, with a mean square error of ∼10−3. Hence, this method compresses the droplet contour to merely two numerical parameters via a fully reversible process, a crucial feature for rendering learning algorithms computationally tractable.

A Novel Approach for Stratifying Pulmonary Edema Severity on Chest X-ray via Dual-Mechanic Self-Learning and Bidirectional Multi-Modal Cross-Attention Algorithms

Abstract Accurate assessment of pulmonary edema severity in acute decompensated congestive heart failure (CHF) patients is vital for treatment decisions. Traditional methods face challenges due to the complexity of chest X-ray (CXR) and unstructured radiology reports. We proposed a method combining self-supervised learning and multimodal cross-attention to address these challenges. Dual-mechanic self-supervised pre-training enhances feature extraction using contrastive learning between text and image features, and generative learning between images. A bidirectional multi-modal cross-attention model integrates image and text information for fine-tuning, improving model performance. Four CXR datasets consisting of 519, 437 images were used for pre-training; 1200 randomly selected image-text pairs were used for fine-tuning and partitioned into train, validation, and test sets at 3: 1: 1. Ablation studies for pre-training and fine-tuning approaches indicated their practicality as evidenced by the optimal macro F1 score of 0.667 and optimal macro-AUC of 0.904. It also outperformed other state-of-the-art multi-modality methods. The novel approach could accurately assess pulmonary edema severity, offering crucial support for CHF patient management.

ORASIS-MAE Harnesses the Potential of Self-Learning from Partially Annotated Clinical Eye Movement Records

Self-supervised learning (SSL) has gained significant attention in the past decade for its capacity to utilize non-annotated datasets to learn meaningful data representations. In the medical domain, the challenge of constructing large annotated datasets presents a significant limitation, rendering SSL an ideal approach to address this constraint. In this study, we introduce a novel pretext task tailored to stimulus-driven eye movement data, along with a denoising task to improve the robustness against simulated eye tracking failures. Our proposed task aims to capture both the characteristics of the pilot (brain) and the motor (eye) by learning to reconstruct the eye movement position signal using up to 12.5% of the unmasked eye movement signal patches, along with the entire REMOBI target signal. Thus, the encoder learns a high-dimensional representation using a multivariate time series of length 8192 points, corresponding to approximately 40 s. We evaluate the learned representation on screening eight distinct groups of pathologies, including dyslexia, reading disorder, and attention deficit disorder, across four datasets of varying complexity and size. Furthermore, we explore various head architecture designs along with different transfer learning methods, demonstrating promising results with improvements of up to approximately 15%, leading to an overall macro F1 score of 61% and 61.5% on the Saccade and the Vergence datasets, respectively. Notably, our method achieves macro F1 scores of 64.7%, 66.1%, and 61.1% for screening dyslexia, reading disorder, and attention deficit disorder, respectively, on clinical data. These findings underscore the potential of self-learning algorithms in pathology screening, particularly in domains involving complex data such as stimulus-driven eye movement analysis.

Collusion by mistake: Does algorithmic sophistication drive supra-competitive profits?

A burgeoning literature shows that self-learning algorithms may, under some conditions, reach seemingly-collusive outcomes: after repeated interaction, competing algorithms earn supra-competitive profits, at the expense of efficiency and consumer welfare. This paper offers evidence that such behavior can stem from insufficient exploration during the learning process and that algorithmic sophistication might increase competition. In particular, we show that allowing for more thorough exploration does lead otherwise seemingly-collusive Q-learning algorithms to play more competitively. We first provide a theoretical illustration of this phenomenon by analyzing the competition between two stylized Q-learning algorithms in a Prisoner’s Dilemma framework. Second, via simulations, we show that some more sophisticated algorithms exploit the seemingly-collusive ones. Following these results, we argue that the advancement of algorithms in sophistication and computational capabilities may, in some situations, provide a solution to the challenge of algorithmic seeming collusion, rather than exacerbate it.

An Effective Composite Investment Quantitative Strategy with Machine Learning Method

In recent years, significant breakthroughs in AI and machine learning have garnered international attention. One such advancement is the development of key algorithms, including "machine self-learning and self-coding algorithms." These algorithms enable machines to continually learn from data and adjust parameters autonomously, without human intervention. Additionally, machines engage in "reinforcement learning," where algorithms achieve set goals through constant trial and error in specific environments. This has allowed current AI systems to adapt to more complex settings and learn more intricate content. In terms of data processing, AI has also made significant strides, particularly with "Recurrent Neural Networks" enabling current AI to handle sequential data more effectively like neural networks, natural language texts, and speech. Consequently, applying these technologies in the financial investment sector has become a hot topic in the industry. This paper outlines the limitations of traditional methods and proposes a new quantitative composite investment strategy. Our main approach involves designing an intelligent investment advisory model capable of self-learning, continuous self-improvement, and market adaptation. Our process includes analyzing industries, collecting extensive stock and securities data, and eliminating noise data. We then construct a model and assess its stability. Finally, accuracy is determined through an indicator. The results of our method show a high degree of similarity between the predicted stock prices and actual stock prices, indicating the effectiveness of our approach. Thus, our method offers a valuable reference for future applications.

Efficient characterizations of multiphoton states with an ultra-thin optical device

Metasurface enables the generation and manipulation of multiphoton entanglement with flat optics, providing a more efficient platform for large-scale photonic quantum information processing. Here, we show that a single metasurface optical device would allow more efficient characterizations of multiphoton entangled states, such as shadow tomography, which generally requires fast and complicated control of optical setups to perform information-complete measurements, a demanding task using conventional optics. The compact and stable device here allows implementations of general positive operator valued measures with a reduced sample complexity and significantly alleviates the experimental complexity to implement shadow tomography. Integrating self-learning and calibration algorithms, we observe notable advantages in the reconstruction of multiphoton entanglement, including using fewer measurements, having higher accuracy, and being robust against experimental imperfections. Our work unveils the feasibility of metasurface as a favorable integrated optical device for efficient characterization of multiphoton entanglement, and sheds light on scalable photonic quantum technologies with ultra-thin optical devices.

Adaptive AI in precision agriculture: A review: Investigating the use of self-learning algorithms in optimizing farm operations based on real-time data

Adaptive AI in precision agriculture: A review: Investigating the use of self-learning algorithms in optimizing farm operations based on real-time data

A Comprehensive Study on Self-Learning Methods and Implications to Autonomous Driving.

As artificial intelligence (AI) has already seen numerous successful applications, the upcoming challenge lies in how to realize artificial general intelligence (AGI). Self-learning algorithms can autonomously acquire knowledge and adapt to new, demanding applications, recognized as one of the most effective techniques to overcome this challenge. Although many related studies have been conducted, there is still no comprehensive and systematic review available, nor well-founded recommendations for the application of autonomous intelligent systems, especially autonomous driving. As a result, this article comprehensively analyzes and classifies self-learning algorithms into three categories: broad self-learning, narrow self-learning, and limited self-learning. These categories are used to describe the popular usage, the most promising techniques, and the current status of hybridization with self-supervised learning. Then, the narrow self-learning is divided into three parts based on the self-learning realization path: sample self-learning, model self-learning, and self-learning architecture. For each method, this article discusses in detail its self-learning capacity, challenges, and applications to autonomous driving. Finally, the future research directions of self-learning algorithms are pointed out. It is expected that this study has the potential to eventually contribute to revolutionizing autonomous driving technology.

Towards Autonomous Programming of Micro-Assembly Robotics

Due to the strive towards miniaturized systems and the growing field of optical technologies, micro-assembly is becoming increasingly important. Micro-assembly is characterized by challenging processes that require sub-micron level positioning accuracy regardless of modeling and calibration errors in the manipulator system. Automating these processes requires not only profound expertise about the process itself but also highly skilled personnel for programming the micro-assembly robot since current interfaces lack intuitive programming methods and simulation capabilities. In this paper, we outline a roadmap towards autonomous programming by combining intuitive programming approaches with intelligent and self-learning algorithms. Following this roadmap, the user will be supported progressively by autonomous and intelligent sub-processes until the machine can finally program itself autonomously. Based on a systematic review of the current state of automated micro-assembly and simulation frameworks, we show the capabilities of current approaches and identify key enablers for an autonomous assembly. From these enablers, we derive modules building our proposed framework. Central aspects are the development of a holistic simulation and a data management, which include not only the robot with its sensor systems but also assembly-components. These form the foundation for offline programming and the usage of machine learning algorithms. In order to facilitate future research, we propose the utilization of the Robot Operating System 2 framework (ROS2) as a basis for autonomous programming adhering the principles of open-source and enabling seamless integration.

ОБЗОР АДАПТИВНЫХ СИСТЕМ УПРАВЛЕНИЯ ДЛЯ БЕСПИЛОТНЫХ МЕХАТРОННЫХ ПЛАТФОРМ

The article considers modern approaches to adaptive control for unmanned mechatronic platforms, including ground, air and underwater devices. Various methods of adapting control systems to changing operating conditions are analyzed, such as the use of predictive models, neural networks and self-learning algorithms. Particular attention is paid to the problems of control under uncertainty and limited computing resources. The advantages and limitations of using adaptive systems in real-life sce-narios are presented, and the prospects for their further development are discussed.

Labor law and artificial intelligence: points of contact and differences

The advent of high-tech and self-learning AI algorithms is setting off an unprecedented transformation of social production processes that will fundamentally affect the entire world of labor. According to the author, the introduction of AI into the world of labor will undoubtedly lead to a temporary surge in technological unemployment, but in the long term, new technologies will create more jobs in new sectors of the economy. The impact of AI on unemployment is context-specific and should be subject to government regulation. The author points out that as a result of the introduction of AI algorithms in the world of labor, arise not only the traditional problems of strengthening the economic power of the employer, discrimination or unauthorized collection of personal data, but set a big complex of legal problems related to the responsibility of the employer for decisions that he himself cannot control. Therefore, for labor law, the most important task is to eliminate the discrepancies between the current model of legal regulation of labor relations and the risks of introducing AI into decision-making processes of hiring and controlling employees. The science of labor law should develop relevant approaches to a reasonable limitation of the use of AI, taking into account the peculiarities of modern algorithmic technologies, production prospects, legal and social risks. The author criticizes proposals for endowing AI with legal personality and the possibility of delegating responsibility for the implementation of employer functions to algorithms. According to the author, AI cannot have legal personality in labor relations (as well as in legal relations of any other type), since the functioning of AI is carried out through the datification of all their participants without the goal of achieving a socially significant result and establishing interaction between subjects of law regarding the satisfaction of their needs. AI is a means of automating labor processes, a digital interface for interaction between elements of the production system. The author states that AI in legal reality can exist only as an object of law. The author proposes to fix in the labor legislation the presumption of responsibility of the employer for the decisions made by AI, regardless of the originally programmed algorithms, even if they were changed by AI as a result of machine learning.

A Study on Energy-Efficient Wireless Sensor Network based on Machine Learning Techniques

Wireless sensor networks (WSNs) are advantageous when there is no existing infrastructure (such as in military applications, emergency relief efforts, etc.) and it is necessary to develop a network at a low cost. A predetermined routing protocol or intrusion detection system is not available to Wireless Sensor Networks because they are dynamic by nature and need separating the network's nodes to do this. Because nodes in the majority of WSN applications are mobile and rely on battery capacity and the availability of restricted resources, energy consumption is an important research area for carrying out a variety of activities in WSNs. Self-learning algorithms that function without scripting or human involvement can be effectively used to report this problem depending on the applications need. This study investigated different ML-based WSN systems and exploring the ML techniques for energy efficiency along with some open issues.

Self-learning Agents for Recommerce Markets

Nowadays, customers as well as retailers look for increased sustainability. Recommerce markets – which offer the opportunity to trade-in and resell used products – are constantly growing and help to use resources more efficiently. To manage the additional prices for the trade-in and the resale of used product versions challenges retailers as substitution and cannibalization effects have to be taken into account. An unknown customer behavior as well as competition with other merchants regarding both sales and buying back resources further increases the problem’s complexity. Reinforcement learning (RL) algorithms offer the potential to deal with such tasks. However, before being applied in practice, self-learning algorithms need to be tested synthetically to examine whether they and which work in different market scenarios. In the paper, the authors evaluate and compare different state-of-the-art RL algorithms within a recommerce market simulation framework. They find that RL agents outperform rule-based benchmark strategies in duopoly and oligopoly scenarios. Further, the authors investigate the competition between RL agents via self-play and study how performance results are affected if more or less information is observable (cf. state components). Using an ablation study, they test the influence of various model parameters and infer managerial insights. Finally, to be able to apply self-learning agents in practice, the authors show how to calibrate synthetic test environments from observable data to be used for effective pre-training.

Intelligent information systems of the banking sector: General characteristics and information environment

Due to the rapid development of digitalization and information technology, the study of intelligent information systems (IIS) in the banking sector is becoming an urgent task. Intelligent information systems are able to optimize banking processes, increase security, improve the quality of customer service, mitigate risks, and optimize internal processes of financial stability management. The purpose of this study is to reveal the potential and impact of IIS on banking management processes; to study their capabilities. To achieve this goal, this study used an analytical approach, in particular, methods of information and morphological data analysis, as well as the method of generalization, which allowed the identification of key aspects, features, and properties of intelligent information systems of the banking sector, and to provide a generalized structure of their functioning by the relevant processes. This article generalizes a group of intelligent information systems of the banking sector (IISBS), their general features and properties, proposes an innovative architecture of financial management support, and identifies their advantages in comparison with existing intelligent systems. The study proves that intelligent information systems of the banking sector are endowed with hybrid data analytics provided by deep learning methods using self-learning algorithms; are able to assess possible risks and plan strategies for their resolution; recognize unauthorized entries and suspicious transactions; thanks to virtual assistants, are suitable for robotizing management processes; and visually present the results of analysing large amounts of data in real time. The research conducted in this paper shows that the introduction of intelligent information systems in the banking sector is of high practical value, as it provides interactivity and personalization for customers, online interaction, and support in solving problems through various communication channels

Battery and Hydrogen Energy Storage Control in a Smart Energy Network with Flexible Energy Demand Using Deep Reinforcement Learning

Smart energy networks provide an effective means to accommodate high penetrations of variable renewable energy sources like solar and wind, which are key for the deep decarbonisation of energy production. However, given the variability of the renewables as well as the energy demand, it is imperative to develop effective control and energy storage schemes to manage the variable energy generation and achieve desired system economics and environmental goals. In this paper, we introduce a hybrid energy storage system composed of battery and hydrogen energy storage to handle the uncertainties related to electricity prices, renewable energy production, and consumption. We aim to improve renewable energy utilisation and minimise energy costs and carbon emissions while ensuring energy reliability and stability within the network. To achieve this, we propose a multi-agent deep deterministic policy gradient approach, which is a deep reinforcement learning-based control strategy to optimise the scheduling of the hybrid energy storage system and energy demand in real time. The proposed approach is model-free and does not require explicit knowledge and rigorous mathematical models of the smart energy network environment. Simulation results based on real-world data show that (i) integration and optimised operation of the hybrid energy storage system and energy demand reduce carbon emissions by 78.69%, improve cost savings by 23.5%, and improve renewable energy utilisation by over 13.2% compared to other baseline models; and (ii) the proposed algorithm outperforms the state-of-the-art self-learning algorithms like the deep-Q network.

Rights of investors in the context of algorithmic Artificial Intelligence technologies and automatization

This study explores the profound impact of digital technologies on economic processes, particularly in the realm of investments. Financial organizations are leveraging digital platforms and algorithmic artificial intelligence (AI) to provide enhanced services to consumers. AI-powered algorithms surpass humans in processing vast amounts of information, leading to improved price discovery and reduced transaction costs. Robo-advisors, empowered by AI, have become pivotal players in financial markets, offering personalized investment portfolios based on risk profiles and objectives. While algorithmic AI technologies eliminate human errors and streamline decision-making, they also introduce new risks. Flaws in algorithm construction, intentional errors, cybersecurity vulnerabilities, and the regulatory challenges of supervising self-learning algorithms are concerns. Additionally, the rise of blockchain platforms has transformed cryptocurrencies and digital tokens into new tradable assets with their own protection challenges. Researchers emphasize that the wealth of data and correlations facilitated by AI can have unintended consequences, and the development of proper supervision and control mechanisms is vital. As the financial landscape continues to evolve, it becomes crucial for investors and regulators to navigate the complexities of digital technologies and address the emerging Global Technology Risk (GTR).

Algorithmic pricing

The rise of algorithmic pricing has transformed perfect price discrimination from a theoretical concept into a real possibility. Through self-learning pricing algorithms, a strategy can be developed that approximates consumers’ reservation prices with ever-improving accuracy. This paper analyzes algorithmic pricing from a law and economics perspective to identify the efficiency and equity effects that the practice could cause and determine to which extent it is regulated under the current legal framework. This paper finds that under competitive market conditions, algorithmic pricing could be welcomed from an efficiency perspective, but from an equity and ethical perspective serious concerns need to be raised. If these concerns are to be taken seriously, the legal framework provides only a partially functional approach to address algorithmic pricing. Additional appropriate remedies are, therefore, needed to protect consumers adequately and effectively against exploitation that reduces their welfare.

Survey of Optimization Algorithms in Modern Neural Networks

The main goal of machine learning is the creation of self-learning algorithms in many areas of human activity. It allows a replacement of a person with artificial intelligence in seeking to expand production. The theory of artificial neural networks, which have already replaced humans in many problems, remains the most well-utilized branch of machine learning. Thus, one must select appropriate neural network architectures, data processing, and advanced applied mathematics tools. A common challenge for these networks is achieving the highest accuracy in a short time. This problem is solved by modifying networks and improving data pre-processing, where accuracy increases along with training time. Bt using optimization methods, one can improve the accuracy without increasing the time. In this review, we consider all existing optimization algorithms that meet in neural networks. We present modifications of optimization algorithms of the first, second, and information-geometric order, which are related to information geometry for Fisher–Rao and Bregman metrics. These optimizers have significantly influenced the development of neural networks through geometric and probabilistic tools. We present applications of all the given optimization algorithms, considering the types of neural networks. After that, we show ways to develop optimization algorithms in further research using modern neural networks. Fractional order, bilevel, and gradient-free optimizers can replace classical gradient-based optimizers. Such approaches are induced in graph, spiking, complex-valued, quantum, and wavelet neural networks. Besides pattern recognition, time series prediction, and object detection, there are many other applications in machine learning: quantum computations, partial differential, and integrodifferential equations, and stochastic processes.

AI impacts on supply chain performance : a manufacturing use case study

The integration of cross-company activities to form global supply chains (SC) has several benefits, including reducing costs, minimizing energy and resource waste, and promoting relationships for improving all network actors. However, as the number of tiers of suppliers and customers increases, monitoring processes and identifying problems becomes more challenging, which can threaten the continuity of the SC. To address this issue, the EU knowlEdge project proposes using artificial intelligence (AI) solutions that are distributed, scalable, and collaborative to enable automatic monitoring and learning in the SC. This approach replaces rigid organization with flexible networks that leverage self-learning algorithms and automatic value creation, thereby facilitating knowledge sharing. The project unifies technologies from various domains, including AI, data analytics, edge, and cloud computing, into a software architecture that offers a systemic solution rather than an incremental improvement. This architecture enhances SC performance, including adaptability and autonomy, and enables industry to adopt adaptive strategies. The platform’s functionality is tested in manufacturing, where it will improve production monitoring and planning and enable human intervention and learning. The AI application is expected to increase performance on various business and production indicators, which will also have an impact beyond the factory floor. With this approach, managers can respond quickly to changing customer requirements, while deviations in planned processes can be addressed more effectively. Additionally, the research conducted by the project will provide insights into future management and learning in SC.