This study examines the nexus between selected cryptocurrencies represented by Bitcoin and Ethereum and the expected market volatility denoted by the VIX index. We use daily data for the period of 06/01/2021 – 07/07/2023. Using Hong (2001) and Hong et al. (2009) Granger causality tests we find no causality in mean, however, we find strong bidirectional causality in variance for the variables at the 5% significance level across all lag lengths. At the time-varying causality analysis, we find BTC causes VIX in the 1st lag. The analysis findings have important implications for portfolio managers and potential investors. Although no direct causal relationship was found between Bitcoin and VIX, or Ethereum and VIX in terms of returns, the strong bidirectional causality in variance suggests that changes in volatility can affect related assets. This information can be utilized by portfolio managers to effectively assess mandates and manage portfolio risk.
The development of autonomous vehicles has been a rapidly growing field, with increasing interest in the technology and its potential applications. However, there are currently no commercially available vehicles that can be classified as level 3 or higher on the Society of Automotive Engineers’ autonomy scale. One key aspect of advancing the technology is the optimization of the verification and validation process through the use of simulators. This paper presents an in-depth analysis of the architecture of an autonomous vehicle, developed and tested using Simulink. The focus of the paper is to provide a comprehensive understanding of the path planning phase in autonomous vehicles, which is a crucial aspect of the vehicle’s navigation and decision-making capabilities. Through this research, we aim to gain a deeper understanding of the underlying decision-making processes of autonomous vehicles.
Nowadays, computer games have gained immense popularity, captivating gamers who can easily spend hours engrossed in their laptops. However, not every game has the ability to keep players emotionally engaged. To achieve this, a game requires a unique storyline and well-developed characters, each with their own distinct traits. One genre that tends to be highly addictive is the "survival of the fittest" concept, where a player's success depends on their chosen gameplay strategy. This article explores the utilization of multivariate normal distribution and Poisson distribution for spawning items in "Battle Royale" games. The term "Battle Royale" refers to a specific genre of video games that has gained significant traction in recent years. These games typically involve a large number of players competing simultaneously within a single game session, often ranging from tens to hundreds of participants. The primary objective for players in battle royale games is to survive and outlast their opponents, creating an intense and competitive gaming experience.
At present, LDA topic model, Word2Vec model and knowledge graph have been used to recommend doctors for online consultation, and have achieved certain results. However, the single word vector model will lead to some deficiencies in the recommendation results. In online consultation, patients have short questions and doctors have long answers, which is difficult to deal with uniformly. Therefore, this paper combines Sentence BERT sentence vector model, ELMO word vector model and doctor attribute model to build a set of combined recommendation model for intelligent doctor recommendation. In the empirical study, this paper chooses a large number of test patients as the research object. The results show that compared with the single word vector model, the combined recommendation model proposed in this paper is better in the accuracy rate, mean reciprocal rank and coverage rate of doctor recommendation.
Several probabilistic decoders of second-order Reed-Muller codes are considered. These decoders share a common construction idea based on the differentiation of polynomials. The properties of these decoders are considered. It is shown that it is possible to modify these decoders for the case of any odd Galois field of cardinality more than 3, but it is theoretically impossible to justify the correctness of such algorithms.
As real-time video processing applications grow in complexity, they demand higher performance. Achieving such a performance must involve a delicate balance between design constraints and optimization of performance criteria. A vital aspect of this balance is the integration of application-specific accelerator designs to boost computational efficiency. To illustrate this, we applied Laplacian High-Pass filtering operations on real-time video signals across three hardware platforms an ARM processor, an ARM+FPGA-based SoC, and a single-core Intel i7 processor. We further analyzed these platforms’ price-performance ratios. Our research revealed that the ARM+FPGA-based SoC executed the filtering algorithms 23.124 times faster than the ARM processor and 1.969 times faster than the Intel i7 processor. Additionally, the ARM+FPGA-based SoC also showed the highest price-performance efficiency. To offer readers a more visual understanding, we include a resource utilization graph for the SoC hardware accelerator development board, thus demonstrating the efficiency of each platform tested.
The strengths of fuzzy logic and deep learning are combined in this paper to propose a hierarchical fused fuzzy deep neural network model for data classification. By combining linguistic expertise from fuzzy logic with learnt representations from deep neural networks, the suggested approach seeks to improve classification accuracy and interpretability. Input, fuzzy logic, feature extraction, fusion, and decision layers are just a few of the many layers that make up the model. Through linguistic variables, membership functions, and fuzzy rules, the fuzzy logic layer captures uncertainty and imprecision in the incoming data. In order to extract high-level representations from the data, the feature extraction layer makes use of deep neural networks. The learnt representations from the feature extraction layer are combined with the fuzzy inference outcomes from the fuzzy logic layer in the fusion layer. The combination of language and learnt elements is made possible by this fusion process, which results in a better representation of the input material. The final categorization is subsequently made by the decision layer using the combined data. The efficiency of the suggested model is demonstrated by tests performed on the CIFAR-10 dataset. Comparing the hierarchical fused fuzzy deep neural network to conventional methods and individual techniques, it provides enhanced classification accuracy. The model is capable of handling challenging and realistic classification tasks thanks to the integration of fuzzy logic with deep learning. The suggested model's mix of fuzzy logic with deep learning demonstrates excellent potential for tackling difficult issues in a variety of fields. The combined method improves accuracy, resilience, and interpretability, opening up new opportunities for the development of both artificial intelligence and data classification as a whole.
The article studies the development of an electronic course as an element of the suggested interactive adaptive training system. The authors define the main structural elements of the system and consider their mutually relationships. The article describes the training algorithm providing adaptivity to learners' needs and capabilities. In the frame of the suggested adaptive training system, the paper demonstrates some examples of interactive tools which help students to learn the course materials better. In particular, the article considers a set of developed interactive programs for practices in studying economic and mathematical methods and models.
Software development practices have evolved rapidly, seeking more flexible and reusable architectures. Microservices technology is an example of this, with its principles of technological independence and heterogeneity. These principles can be extended to the concept of micro-frontends, which proposes to divide the interface, allowing each team to work independently. This article presents an architecture that uses components developed in multiple frontend frameworks, aiming to make the workflow more independent and modular. The proposed architecture was evaluated in a prototype developed at the Animal Health Defense Platform of Rio Grande do Sul, which successfully integrated components from different frameworks.
The use of deep neural networks in information retrieval significantly improves its effectiveness, but negatively affects the performance of the process. To deal with this, we propose a new ranking model that uses the deep neural network of the "Transformer" architecture (in particular, BERT) for efficient information retrieval. In accordance with the proposed approach, contextualized vector representations are extracted from documents during indexing, after which these representations are clustered for each independent token. The resulting clusters reflect different meanings of the words and are indirectly used as inverted index keys. The values represent the documents in which these contextualized word meanings occur, along with the distances from each document to the contextualized embedding. Thus, after the indexing process, we obtain an index containing pre-calculated distances between the contextualized meanings of dictionary elements and documents. This approach helps us avoid the performance overhead of calculating distances online. At the search stage, the query is transformed into a set of contextualized vectors representing each query token, which allows us to use these vectors to retrieve most semantically close neighbor-tokens and use them to extract relevant documents from the index. This way of searching for contextualized embeddings consumes less memory and is more performant due to the use of an inverted index.
Attacks on the integrity of Unmanned aerial vehicles (UAVs) onboard software are on the rise. Such attacks are very critical and can lead to man-made incidents. There are several approaches for modeling these attacks. This article explores the interaction between an attacker and a defender. The authors have developed a tree model for attacking firmware integrity. Using this attack-defense tree, a game-theoretic risk assessment scheme is formulated. The effectiveness of the proposed scheme was evaluated on the example of a firmware integrity attack and conclusions were shown confirming the viability of the proposed method.
In recent years, the use of contrastive learning methods for time series classification has shown promising results. State-of-the-art approaches leverage self-supervised contrastive learning representations that encode the underlying series’ features. On top of the latent representation, classification is addressed as a downstream task. While the effectiveness of contrastive models in univariate series classification is established, their portability towards a multivariate scenario is still debated. In this work, we explore the use of contrastive time series representations on multivariate data. We compare the performance of state-of-the-art contrastive models on 30 benchmark datasets and explore the use of an early fusion network to combine the input dimensions. The preliminary results show that incorporating a preliminary stage of information fusion is beneficial to improve the performance of state-of-the-art contrastive time series classifiers.
Generative artificial intelligence (GenAI) in general, and large language models (LLMs) in particular, are highly fashionable.As they have the ability to generate coherent output based on prompts in natural language, they are promoted as tools to free knowledge workers from tedious tasks such as content writing, customer support and routine computer code generation.Unsurprisingly, their application is also attractive to professionals in the research domain, where mundane and laborious tasks, such as literature screening, are commonplace.We evaluate Vertex AI 'text-bison', a foundational LLM model, in a real-world academic scenario by replicating parts of a popular systematic review in the information management domain.By comparing the results of a zero-shot LLM-based approach with those of the original study, we gather evidence on the suitability of state-of-theart general-purpose LLMs for the analysis of scientific content.We show that the LLM-based approach delivers good scientific content analysis performance for a general classification problem (ACC = 0.9), acceptable performance for a domain-specific classification problem (ACC = 0.8) and borderline performance for a text comprehension problem (ACC ≈ 0.69).We conclude that some content analysis tasks with moderate accuracy requirements may be supported by current LLMs.As the technology will evolve rapidly in the foreseeable future, studies on large corpora, where some inaccuracies are tolerable, or workflows that prepare large data sets for human processing, may increasingly benefit from the capabilities of GenAI.
This work presents an optimal design of a 4x4 Butler Matrix (BM), which serves as key component in an antenna system designed for 5G applications operating at 60 GHz. The design is developed using the Advanced Design System (ADS) software provided by Keysight Technologies. The proposed design is implemented on Rogers (RO3003), characterized by loss tangent and dielectric constant of 0.001 and 3, respectively. The Butler 4x4 matrix configuration uses 4 input ports, which allows 4 homogeneous orthogonal beams to be generated at specific angles. The design aims to enhance the beamforming capabilities and overall performance of a smart antenna system, contributing to the advancement of 5G wireless communication technologies.
To meet the increasing demand for data transmission in future spacecraft, a structural revision is necessary. The current limitations of radio frequency communication systems, both technologically and regulatory, have created a bottleneck in data transmission to the ground segment. To address these challenges, the integration of free space optics communication within terrestrial space networks proves crucial. This approach offers several advantages, including significantly enhanced data rates compared to traditional RF technology, cost-effectiveness, improved security, inter-satellite free space communication, and the utilization of laser beams as optical signal carriers for establishing satellite-ground and ground-to-satellite links. Consequently, there is a requirement for high-speed and energy-efficient systems to serve as a foundation for transmitting large volumes of video and audio data. Nano Satellites, including the CubeSat platform, possess advanced technical capabilities that make them essential components in the space sector. They operate in Low-Earth-Orbit and are designed with cost-effectiveness and diverse communication topologies in mind. This research focuses on developing the output parameter indicators for the existing CubeSat platform’s optical communication transceiver subsystem. The objective is to enhance the performance of this communication methodology by introducing a 3D optical switch and a laser beam-controlled optical transponder as structural subsystems within a 2U CubeSat. The study encompasses their functional performance, structural parameters, and their application within the Low Earth Orbit satellite network topology.
Currently, social networks are confidently integrated into the lives of modern people. These networks are a rich source of data. In the context of a large amount of information presented here, the task of intellectual analysis of the content of a social network arises. This article highlights the possibility of using social media information for educational purposes, as well as reviewing relevant research in the field of applying Data Mining methods for intellectual analysis of educational data. The article also shows the possibility of using data clustering methods using k-means to identify the content of educational materials distributed by students in a social network. This data can be used by the teachers of the educational institution to improve educational content and improve its quality.
This paper proposes heart rate variability analysis based on ECG data of the cardiovascular system in normal and pathological conditions by binary classification classical methods: Naive Bayes, Nearest Neighbors, Decision Tree, Random Forest, Gradient Boosting and Logistic Regression. The mean classification quality metrics such as the accuracy, the recall, the precision, the F1-score and the AUC score were obtained as a 10-fold cross-validation test. The electrocardiographic records of patients with normal sinus rhythm (the normal group), chronic heart failure of II and III classes according to the NYHA classification and ventricular extrasystole (the pathology group) from the open research resource PhysioNet became the data of the research. According to the results of the HRV analysis using binary classification methods, the most informative indicators for the cardiovascular pathology detection were determined. The constructed models were tested on the ECG of the Krasnoyarsk Territory patients and the classification quality of each model was evaluated. The best results were obtained by Logistic Regression, k-Nearest Neighbors, Gradient Boosting and Random Forest (indicators: SDANN, SDNN and IRRR).
This paper presents a path loss and coverage analysis study for candidate frequency bands in mmWave region for 5G. We performed numerical analysis over a sample outdoor region in our university campus and compared the coverage regions for different data rates at different frequency bands, namely 28, 60 and 73 GHz. Our results reveal that all frequency bands provide considerable coverage of almost the entire line-of-sight (LOS) region at low data rates. However, when the data rate requirement is increased, the coverage region with 73 GHz becomes approximately 10 times smaller than low data rate applications.
E-government is defined as the effective integrated use of all information and communication technologies to facilitate the daily administrative operations of government sectors. The revolution in digital communication technologies has put more pressure on many public sector institutions to transform their operations into the world of electronic business. This is what is known as e-government. E-government is defined as the effective integrated use of all information and communication technologies. People are recognised by the use of their fingerprints, faces, hand geometries, and iris prints, among other biometric characteristics. This research intends to use iris recognition to identify people through the image of the iris and apply a programme using Python and the deep learning method of a convolutional neural network (CNN) to match the printed iris images of people, where the images match the images of people and show the name of the person who owns this photo. This research also intends to use the deep learning method to match the printed iris images of people. This is a contemporary system that aids in the identification of individuals and replaces e-government systems with smart systems that incorporate artificial intelligence into their operations to deliver the best services possible to residents. How much does the Jordanian government want to employ biometric characteristics, such as the iris of the eye, which is regarded as part of the biometric features, in all of the electronic government services.
Summarizing conversations like meetings, email threads or discussion forums poses relevant challenges on how to model the dialogue structure. Existing approaches mainly focus on premise-claim entailment relationships while neglecting contrasting or uncertain assertions. Furthermore, existing techniques are abstractive, thus requiring a training set consisting of humanly generated summaries. With the twofold aim of enriching the dialogue representation and addressing conversation summarization in the absence of training data, we present an extractive conversation summarization pipeline. We explore the use of contradictions and neutral premise-claim relations, both in the same document or in different documents. The results achieved on four datasets covering different domains show that applying unsupervised methods on top of a refined premise-claim selection achieves competitive performance in most domains.
