Text generation is an important method to generate high quality and available product description from product title. For the product description generation for online E-commerce application, the main problem is how to improve the quality of generated text. In other words, how we judge the quality of text. If all texts are already positive and available, then we find it impossible to manually judge which text is the better text for a product. So if we cannot judge which is a better text manually, we cannot improve the quality of generated text. In E-commerce, product description is to attract shoppers and improve sales. So we design a method to improve the quality of generated text based on user buying behaviour. Online result shows that our approach improve the sales of products by improving the text quality.

Advances in Internet of Things (IoT) technologies have resulted in a significant surge in the utilization of sensor devices across diverse domains for environmental sensing and monitoring. The applications of IoT sensor devices in environmental monitoring span a wide range, including the surveillance of biodiverse areas such as peatlands, forests, and oceans, as well as air quality monitoring, commercial agriculture, and the safeguarding of endangered species. This paper provides a long term evaluation of IoT sensors data quality in environmental monitoring networks, particularly focusing on peatland regions. IoT sensors have the capacity to provide high resolution spatiotemporal dataset in environmental monitoring networks. Sensor data quality plays significant role in increasing the adoption of IoT devices for environmental data gathering. However, due to the nature of deployment (i.e., in harsh and unfavourable weather conditions), coupled with the limitations of low-cost components, IoT sensors are prone to collection of erroneous data, also the nature of peatland ecosystems presents unique challenges in data quality assurance due to their complex and dynamic characteristics. This paper identifies specific challenges and issues related to IoT sensor data quality in different peatland ecotopes. These challenges include sensor placement and calibration, data validation and fusion, environmental interference, and the management of data gaps and uncertainties. To address these challenges, the paper presents and evaluates methods for improving data quality in peatland monitoring networks. These methods encompass advanced sensor calibration techniques, data validation algorithms, machine learning approaches, data processing and data fusion strategies.

This paper describes a method for automatically transforming the structure and characteristics of an image processing dataflow graph for the purpose of improving performance and/or lowering memory utilization as compared to the baseline tools. Embedded image processing applications are often executed on Digital Signal Processors, or their modern equivalent Visual Processor Units. The software usually performs a series of pixel-level operations for basic color conversion, channel extraction and combining, arithmetic, and filtering. These steps can often be efficiently described as a graph. For this reason, standard libraries such as OpenVX are used, which provide a graph-based programming model where the nodes are chosen from a repertoire of common pixel-level operations and the edges represent the flow of images as they progress though the processing stages. Generally speaking, each node is processed sequentially in the order implied by the data dependencies defined by the graph structure, with all intermediate values stored in external memory. In the proposed framework, we developed performance models for both the direct memory access subsystem and the L1 data cache to allow for selection of certain intermediate values to be stored in on-chip scratchpad memory as well as selecting the most appropriate tile size. In this way, we effectively decompose the graph in a way to fuse specific sets of nodes to associate their internal edges with on-chip buffers. Additionally, the tile size is optimized for each fused set of nodes. In this paper, we describe our performance models and approach for graph decomposition and tile size selection. The proposed performance models are accurate to within 2% on average, and the overall approach of graph optimization achieves an average speedup of 1.3 and allows for reduction of average DRAM utilization from 100% to as low as 15%.

Cryptography has become an essential tool in information security, preserving data confidentiality, integrity, and availability. However, despite rigorous analysis, cryptographic algorithms may still be susceptible to attack when used on real-world devices. Side-channel attacks (SCAs) are physical attacks that target cryptographic equipment through quantifiable phenomena such as power consumption, operational times, and EM radiation. These attacks are considered to be a significant threat to cryptography since they compromise the integrity of the algorithm by obtaining the internal cryptographic key of a device by seeing its physical implementation. The literature on SCAs has focused on real-world devices, yet with the growing popularity of sophisticated devices like smartphones, fresh approaches to SCAs are necessary. One such approach is electromagnetic side-channel analysis (EM-SCA), which gathers information by listening to electromagnetic (EM) radiation. EM-SCA has been demonstrated to recover sensitive data like encryption keys and has the potential to identify malicious software, retrieve data, and identify program activity. This study aims to evaluate how well EM-SCA compromises encryption under various application scenarios, as well as examine the role of EM-SCA in digital forensics and law enforcement. Regarding this, addressing the susceptibility of encryption algorithms to EM-SCA approaches can provide digital forensic investigators with the tools they desire to overcome the challenges posed by strong encryption, allowing them to continue playing a crucial role in law enforcement and the justice system. Furthermore, this paper seeks to define the current state of EM-SCA in terms of attacking encryption, the encryption algorithms and encrypted devices that are most vulnerable and resistant to EM-SCA, and the most promising EM-SCA on encryption approaches. This study will provide a comprehensive analysis of EM-SCA in the context of law enforcement and digital forensics and point towards potential directions for further research.

In order to low-frequency stabilize the electric field integral equation (EFIE) when discretized with divergence conforming B-spline based basis and testing functions in an isogeometric approach, we propose a corresponding quasi-Helmholtz preconditioner. To this end, we derive i) a loop-star decomposition for the B-spline basis in the form of sparse mapping matrices applicable to arbitrary polynomial orders of the basis as well as to open and closed geometries described by single-or multipatch parametric surfaces (as an example non-uniform rational Bsplines (NURBS) surfaces are considered). Based on the loopstar analysis, we show ii) that quasi-Helmholtz projectors can be defined efficiently. This renders the proposed low-frequency stabilization directly applicable to multiply-connected geometries without the need to search for global loops and results in betterconditioned system matrices compared to directly using the loopstar basis. Numerical results demonstrate the effectiveness of the proposed approach.

In order to accurately compute scattered and radiated fields in the presence of arbitrary excitations, a lowfrequency stable discretization of the right-hand side (RHS) of a quasi-Helmholtz preconditioned electric field integral equation (EFIE) on multiply-connected geometries is introduced, which avoids an ad-hoc extraction of the static contribution of the RHS when tested with solenoidal functions. To obtain an excitation agnostic approach, our approach generalizes a technique to multiply-connected geometries where the testing of the RHS with loop functions is replaced by a testing of the normal component of the magnetic field with a scalar function. To this end, we leverage orientable global loop functions that are formed by a chain of Rao-Wilton-Glisson (RWG) functions around the holes and handles of the geometry, for which we introduce cap surfaces that allow to uniquely define a suitable scalar function. We show that this approach works with open and closed, orientable and non-orientable geometries. The numerical results demonstrate the effectiveness of this approach.

Manual object identification labelling is laborious, time-consuming and prone to inconsistencies hindering advancements in various computer vision tasks.These inconsistencies can lead to inaccurate models with poor performance. Considering these potential consequences, highlights the importance of addressing labelling challenges for ethical and responsible AI development. To address this our study evaluates several popular platforms for their suitability in tackling these challenges. Roboflow, Makesense.ai, SentiSight.ai, Labelbox and SuperAnnotate are the five different data labelling platforms that have been taken for assessment. The study identifies strengths and weaknesses of each platform in the context of basketball detection using YOLO v8, a deep learning model for object detection, image classification, and image segmentation. Each platform is analysed based on features, ease of use, pricing, and support for image annotation, object detection, and YOLO v8 integration. After analysing these factors, a final recommendation is made, highlighting the platform that demonstrably offers the best balance of features, efficiency, and cost-effectiveness for this specific task. The study helps in deeper exploration of the potential of YOLO v8. It is mainly aimed at assisting the Video Assistant Referees(VARs) for accurate and unbiased decision-making and also empowers the development of AI technology across the domain of sports.

Flood monitoring with satellite images is an effective method of detecting and tracking floods. This approach involves the use of satellite imagery to detect changes in water levels and identify flooded areas. To monitor floods using satellite images, the images are analyzed to detect changes in water levels over time. To detect changes in water levels and identify flooded areas based on a set of predefined criteria, we can train algorithms. Amazon SageMaker geospatial capabilities make it easier for data scientists and machine learning (ML) engineers to build, train, and deploy ML models using geospatial data.  These capabilities also provide pre-trained models. One of the pre-trained models is land cover segmentation model. This land cover segmentation model can be run with a simple API call and can be leverage to analyze changes in the water level.

Predictive monitoring on distributed critical infrastructures (DCI) is the ability to anticipate events that will likely occur in the DCI before they actually appear, improving the response time to avoid the rise of critical incidents. Distributed into a region or country, DCIs such as smart grids or microgrids rely on IoT, edge-fog continuum computing and the growing capabilities of distributed application architectures to collect, transport, and process data generated by the infrastructure. We present a model-agnostic distributed architecture for the inference execution of machine learning window-based prediction models of predictive monitoring applications to be used in this context. This architecture transports the events generated by the DCI using event streams to be processed by a hierarchy of nodes holding predictive models. It also handles the offloading of inferences from resource-scarce devices at lower levels to the resourceful upper nodes. Therefore, the timing requirements for setting predictions before they occur are met.

The actual paper presents an in-depth study and experimental development of a class of rotorcraft, named as x-tilt, that features four tilting rotors. Initially, the equations of motion modeling the aerial robot are presented based on the Euler-Lagrange formulation. The model includes the aerodynamic effects induced by the rotorcraft's relative motion and propellers. For control purposes the aforementioned model is split into a nominal model and lumped disturbance terms, the latter encompassing endogenous and exogenous uncertainties. In this vein, the actual work propose a robust navigation strategy targeting a specific performance profile whose problem is formulated through the model predictive control (MPC) framework. To this end, two schemes are proposed, (i) an integral MPC and a (ii) MP sliding-mode Control (MPSMC). Both control schemes are linked to a extended-state Linear Kalman Filter (ES-LKF) that furnishes the states and lumped disturbance estimates. Moreover, a high-fidelity simulation is presented in detail to validate the effectiveness of the proposed controller within a realistic scenario. We finally present the experimental stage to validate the tilting-rotor configuration as well as the integral MPC.

Lechner, Hauke, and Zoller proposed a parity-encoded spin-embedding scheme for quantum annealing (QA) with all-to-all connectivity to avoid the issue of limited connectivity in near-term QA hardware and to enable the implementation thereof using only geometrically local interactions between spins fabricated on the planar substrate. Nevertheless, the redundant encoding of logical information, i.e., using a large number of spins to embed the logical information, increases the computational cost and reduces the efficiency. In this study, we show through Monte Carlo simulation that this redundant encoding may be exploited to solve the problems of the inefficiency and computational cost of the parity-encoded scheme by incorporating appropriate decoding, namely classical post-processing, of the spins to retrieve the logical information. Our findings open up the possibility of parity-encoded schemes for realizing the QA with near-term quantum technologies.

This paper reviews about the field of data science and the importance of it with the today's rapidly evolving landscape of technology. We start the survey by introducing the fundamental concepts of data science including history and how we collect data and why we need data science for and how we use them effectively to use them for processing, storing, and analyzing such as machine learning, data visualization. And we explore about the different domains we use data science and the challenges, advantage and disadvantages we face while using them. And finally, we are going to discuss about the prospects and implications of data science and how we can use data science to overcome the current and future challenges. The need of careers and the paths of careers are also discussed as they are essential and growing rapidly.

The Collatz conjecture, a longstanding mathematical puzzle, posits that, regardless of the starting integer, iteratively applying a specific formula will eventually lead to the value 1. This paper introduces a novel approach to validate the Collatz conjecture by leveraging the binary representation of generated numbers. Each transition in the sequence is predetermined using the Collatz conjecture formula, yet the path of transitions is revealed to be intricate, involving alternating increases and decreases for each initial value. The study delves into the global flow of the sequence, investigating the behavior of the generated numbers as they progress toward the termination value of 1. The analysis utilizes the concept of probability to shed light on the complex dynamics of the Collatz conjecture. By incorporating probabilistic methods, this research aims to unravel the underlying patterns and tendencies that govern the convergence of the sequence. The findings contribute to a deeper understanding of the Collatz conjecture, offering insights into the inherent complexities of its trajectories. This work not only validates the conjecture through binary representation but also provides a probabilistic framework to elucidate the global flow of the sequence, enriching our comprehension of this enduring mathematical mystery.

Horizontal Federated Learning exhibits substantial similarities in feature space across distinct clients. However, not all features contribute significantly to the training of the global model. Moreover, the curse of dimensionality delays the training. Therefore, reducing irrelevant and redundant features from the feature space makes training faster and inexpensive. This work aims to identify the common feature subset from the clients in federated settings. Banerjee et al. introduced Fed-FiS 1 , and here we propose a hybrid approach known as Fed-MOFS, where Mutual Information and Clustering are used to select local features from each client. In both approaches, the selection of local features is similar, but Fed-FiS uses a scoring function to evaluate the global ranking of each feature, while Fed-MOFS exploits multi-objective optimization to rank the features based on their higher relevance and lower redundancy criteria. We select the feature subset based on the global ranks for learning. Empirically, we evaluated the performance, stability, and efficacy of Fed-FiS and Fed-MOFS on 12 datasets. We compared Fed-FiS and Fed-MOFS with conventional techniques such as ANOVA and RFE and a federated feature selection method called FSHFL. The experimental results demonstrate both Fed-FiS and Fed-MOFS improve the performance of the global model even after 50% reduction in the feature space size. Both Fed-FiS and Fed-MOFS are at least 2× faster than FSHFL. Also we verified the effect of feature selection on the convergence of the global model. The computational complexity of Fed-FiS and Fed-MOFS is O(d2) and O(2d2), respectively, which is lower than state-of-the-art.

This paper is a continuation on my revolutionary theory of solving the pointwise fluid flow approximation model for time-varying queues. Thus, the long-standing simulative approach has now been replaced by an exact solution by using a constant ratio 𝛽 (Ismail's ratio) , offering an exact analytical solution. The stability dynamics of the time-varying 𝑀/𝐸 𝑘 /1 queueing system are then examined numerically in relation to time, 𝛽, and the queueing parameters.

The paper provides a comprehensive overview of Neural Architecture Search (NAS), emphasizing its evolution from manual design to automated, computationally-driven approaches. It covers the inception and growth of NAS, highlighting its application across various domains, including medical imaging and natural language processing. The document details the shift from expert-driven design to algorithm-driven processes, exploring initial methodologies like reinforcement learning and evolutionary algorithms. It also discusses the challenges of computational demands and the emergence of efficient NAS methodologies, such as Differentiable Architecture Search and hardware-aware NAS. The paper further elaborates on NAS's application in computer vision, NLP, and beyond, demonstrating its versatility and potential for optimizing neural network architectures across different tasks. Future directions and challenges, including computational efficiency and the integration with emerging AI domains, are addressed, showcasing NAS's dynamic nature and its continued evolution towards more sophisticated and efficient architecture search methods.

Distributed training of deep learning models on resource-constrained devices has gained significant interest. Federated Learning (FL) and Split learning (SL) have become the most popular way to do this task. Training data-driven deep learning models in FL/SL involves collaboration between several clients while ensuring user privacy. We aim to optimize these techniques by reducing device computation during parallel model training, and also reducing high communication costs due to model or frequent data and gradient exchanges. This paper proposes Efficient Split Learning (ESL), a novel approach addressing these challenges through three key ideas: (1) a keyvalue store for caching and sharing intermediate activations across clients, significantly reducing redundant computations and communication during the training phase, (2) customization of state-of-the-art neural networks for split learning context, and (3) personalized training allowing clients to learn individual models tailored to their specific data distributions. Unlike previous methods, ESL prioritizes performance optimization while minimizing communication and computation overhead. Extensive experimentation on real-world federated benchmarks for image classification and 3D segmentation demonstrates significant improvements over baseline FL techniques: ESL achieves a reduction in computation by 1623x for image classification and 23.9X for 3D segmentation on resource-constrained devices. Additionally, it reduces communication traffic, during training, between clients and the server by 3.92x for image classification and 1.3x for 3D segmentation, while improving accuracy by 35% and 31%, respectively. Furthermore, when compared to the baseline SL approaches, ESL reduces communication traffic during training by 60x and improves accuracy by an average of 34.8%.

The use of the Choquet integral in multicriteria decision making problems has gained attention in the last two decades. Despite of its usefulness, there is the issue of how to define the Choquet integral parameters, called capacity coefficients, specially the ones associated with coalitions of criteria. A possible approach to address this issue is based on unsupervised learning, which aims to define such parameters with the goal of mitigating undesirable effects provided by intercriteria relations. However, current unsupervised approaches present some drawbacks, as there is no guarantee that the parameters are equally prioritized in the learning procedure. In this paper, we propose a novel unsupervised capacity identification approach which ensures a fair learning for all parameters. Moreover, in comparison with the existing methods, our proposal is less complex in terms of optimization, as it is based on a linear formulation. Experimental results in both synthetic and real datasets attest the applicability of our proposal.

The conventional use of the Retrieval-Augmented Generation (RAG) architecture has proven effective for retrieving information from diverse documents. However, challenges arise in handling complex table queries, especially within PDF documents containing intricate tabular structures. This research introduces an innovative approach to enhance the accuracy of complex table queries in RAG-based systems. Our methodology involves storing PDFs in the retrieval database and extracting tabular content separately. The extracted tables undergo a process of context enrichment, concatenating headers with corresponding values. To ensure a comprehensive understanding of the enriched data, we employ a fine-tuned version of the Llama-2-chat language model for summarisation within the RAG architecture. Furthermore, we augment the tabular data with contextual sense using the ChatGPT 3.5 API through a one-shot prompt. This enriched data is then fed into the retrieval database alongside other PDFs. Our approach aims to significantly improve the precision of complex table queries, offering a promising solution to a longstanding challenge in information retrieval.

The reengineering of business processes to meet business targets revolves around translating business models into equivalent system models. The failure of many software products has been attributed due to system models not mirroring the business processes they have been developed for. As a result many approaches have been suggested on how to bridge the gap between business processes and systems. This paper presents a case for the use of Business Process Execution Language for Web Services rather than Use-cases in the bridging this gap..

P versus NP is considered as one of the most fundamental open problems in computer science. This consists in knowing the answer of the following question: Is P equal to NP? It was essentially mentioned in 1955 from a letter written by John Nash to the United States National Security Agency. However, a precise statement of the P versus NP problem was introduced independently by Stephen Cook and Leonid Levin. Since that date, all efforts to find a proof for this problem have failed. Another major complexity class is NP-complete. It is well-known that P is equal to NP under the assumption of the existence of a polynomial time algorithm for some NP-complete. We show that the Monotone Weighted Xor 2-satisfiability problem (MWX2SAT) is NP-complete and P at the same time. Certainly, we make a polynomial time reduction from every directed graph and positive integer k in the K-CLOSURE problem to an instance of MWX2SAT. In this way, we show that MWX2SAT is also an NP-complete problem. Moreover, we create and implement a polynomial time algorithm which decides the instances of MWX2SAT. Consequently, we prove that P = NP.

We discuss an encryption scheme based on reinforcement learning where a ciphertext is given by an optimal path from the message given by the Bellman equation. We then prove that the encryption scheme is CP A-secure. We discuss possible ways to use the photoelectric effect in electric batteries and the construction of a carbon dioxide filter.

I show that a computer cannot have unambiguous thoughts, not even about a number. What we believe computers do is our own convention. It may seem objective because we anchor it in the user interface. But many other conventions are possible, and they yield different computations, equally valid according to the principles of Computer Science. I prove that the alternative computations equally happen when a single computation is carried out, and in principle they can be accessed. I exemplify this with a program that computes the result for a given input, and then decodes it into the results for all other possible inputs. If thinking would be a computation, a computer would have different, possibly opposite thoughts, corresponding to many alternative computations it implements at the same time. I show probabilistically that the human mind does not have this ambiguity. Therefore, even if the human mind can be simulated by a computer, it cannot be reduced to computation.

Cybersecurity ecosystem is an important facet in protecting sensitive information and securing critical infrastructure for countering modern cyber threats.  With the increasing complexity and frequency of security incidents, there is an escalating demand for development of innovative solutions beyond current human capabilities pertaining to cybersecurity measures.  Artificial Intelligence or AI can be utilized in a myriad of areas of cybersecurity.  It emerged as a technological innovation to enhance cyber protection by facilitating faster and real-time threat detection for known and unknown threats, automating processes to minimize human error, and optimal decision-making.  Harnessing the power of AI in cybersecurity creates formidable defense capabilities against the constantly changing cyber threats of future while empowering the cybersecurity personnel with threat intelligence and proactive foresight to safeguard critical assets and confidential information with unparalleled precision and effectiveness.  This research paper aims to investigate the potential of AI-enabled cybersecurity systems and focuses on deducing the benefits of using AI  in enhancing cybersecurity processes for organizations seeking to manage their risk profile.  Through a comprehensive literature review, the wide-ranging applications of AI in cybersecurity have been analyzed such as intrusion detection, predictive simulation, and automated emergency response management.  The study examines the benefits of implementing AI-based cyber defenses such as improved promptness and accuracy in vulnerability assessment and threat management, reduced false positives, and recognize patterns.  The future potential of AI in cybersecurity will take a leap forward in expanding protection mechanisms to evaluate the strengths and weaknesses of attack vectors to prevent an adversarial attack.