A document by Hao Qin . Click on the document to view its contents.

This paper introduces a semantic communications architecture that is focused on the transmission of american sign language (ASL). Initially, a novel system model for image-based semantic communications is presented, which utilizes a variant of the quadrature amplitude modulation (QAM) scheme, named 24-QAM. This modulation scheme is derived from the original 32-QAM constellation by removing 8 peripheral symbols and is proven capable of attaining superior error performance in ASL applications. Additionally, a semantic encoder based on a convolutional neural network (CNN) which effectively utilizes the ASL alphabet is presented. An original dataset is created by superimposing redgreen- blue landmarks and key-points on top of the captured images; hence, enhancing the representation of hand posture. Finally, the training, testing, and communication performance of the proposed system is quantified through numerical results that highlight the achievable gains and trigger insightful discussions.

Monitoring performance in organizations can act as a catalyst for (i) fostering a culture of continuous improvement over-time, and (ii) achieving their overarching business goals. To this aims, organizations employ Key Performance Indicators (KPIs). However, a common pitfall is the collection of excessive data which is partially disconnected from the business goals. In this study, we propose a KPI framework in the form of a model, an associated template, and a checklist with the aim of providing guidance to practitioners for creating sound KPIs. We examine KPI development and monitoring through four case studies in the context of software systems in two organizations. Through these case studies, we evaluate our KPI framework for its utility in practice and gain insights to uncover related challenges.

This paper presents the analysis of thermal management and cooling DC fan,which cools multi-cores (four cores of Cortex A53 and two cores of Cortes R5F) processor system on a chip and associated components mounted on a single board computer of XCZU3EG-1SBVA484E by forced convection, optimizing the design, operational conditions and maximizing the performance.

The emergence of virtualized computing and converged networks has the potential to revolutionize industrial automation systems, offering unprecedented levels of flexibility and scalability that were previously unattainable with traditional infrastructure. Therefore, cloud-fog automation (CFA) platforms are becoming key players not only in coordinating the production task management in upper-layer applications but also in handling real-time control and monitoring by the edge devices. In this article, we discuss how cloud/fog-based virtualization and converged communication networks can allow for flexible deployment of automation use cases and evaluate the impacts introduced by the less deterministic infrastructure from the perspective of control. Four representative use cases are studied using our in-house developed CFA platform to demonstrate the smart coordination of automation in a variety of real-life prototypes. Moreover, the capabilities of the commercial off-the-shelf virtualized computing and converged networks are characterized, which confirms the CFA platform as an essential part of the future industrial infrastructure for hosting heterogenous control applications is not a “daydream”.

Surrogate-assisted evolutionary algorithms (SAEAs) have been widely used to solve computationally expensive optimization problems. The extreme learning machine-based DirectRanker (ELDR) is a single-layer feed-forward neural network surrogate model designed for SAEAs. ELDR estimates the superiority of two solutions and uses random input weights and biases while learning the output weights. However, ELDR requires high computation time as the problem dimensionality and the number of hidden neurons increase, making its applicability to high-dimensional problems difficult. A surrogate model should be computationally efficient and enable rapid fitness estimations. Therefore, this study proposes a fast implementation technique, i.e., a learning method for ELDR (fELDR) that achieves mathematically equivalent learning results with small computational complexity. Additionally, we propose a pointwise score function to reuse the prediction results. The experimental results confirmed the effectiveness of fELDR compared with the original ELDR. The learning results of the proposed fELDR were equivalent to those of the original ELDR while reducing the training time by up to 97%, especially when using a large hidden layer on a large dimensionality problem. Moreover, owing to the reusable prediction results, the computation time of the fELDR-assisted SAEA can be further decreased by 79.5% compared with that of the original ELDR-assisted SAEA. The reduced training time and reusable prediction results of fELDR make it feasible to apply ELDR to high-dimensional optimization problems and use high prediction accuracy with a large number of hidden neurons.

A document by Weihua Zhou . Click on the document to view its contents.

Digital divide refers to the gap between populations at different socio-economic levels with respect to their access to the Internet and communication technologies in general. In this chapter, we summarize the recent developments aimed at bridging such inequality from a technical and economic standpoint. After a preliminary discussion on alternative backhaul paradigms, we focus on space-, air-, and ground-based communication paradigms, as well as their integration. Subsequently, we discuss the affordability of these solutions, as it is primarily the economic aspect which hinders digital inclusion.

This paper offers for the first time a novel method for creating Predictive Associative Memories of Time series (PAMeT), where a full set of time series variables is used to create a machine learning predictive model (global learning), and afterward, only a few temporal variables are used at a shorter time to recall the model on new data (local recall). Inspired by human brain processes, PAMeT-SNN leverages the brain-inspired SNN NeuCube and the concept of spatio-temporal associative memories (STAM). PAMeT-SNN is a 4-dimensional spatio-temporal structure. First, it encodes time series data into spike sequences that reflects on the changes in the data over time and then maps the temporal variables into the SNN using temporal similarity to define the spatial locations of the variables in the SNN. It then learns temporal associations between a full set of time series variables, thereby memorizing these temporal associations as spatio-temporal connections between neurons. These connections are activated when only part of the time series data is used to recall the model on new data. The proposed groundbreaking method is exemplified with PAMeT-SNN for predictive modeling on two distinct case study time series data sets: trade dynamics and commodity prices. In these case studies, the method effectively captures the intricate data dynamics, enabling accurate forecasting of future values using a minimal set of variables. The method has the potential to be applied in diverse domains.

This paper proposes a specialized artificial neural network layer named SimilarNet, designed to compare two feature vectors while considering the positions of their elements to produce an output in a vector format. By leveraging the advantages of the cosine similarity and concatenation layer, SimilarNet can effectively compare two vectors, enabling the construction of trained comparison models using multidimensional activation functions. In addition, SimilarNet can realize 1:1 comparisons of data with different shapes, unlike cosine similarity.

Digital Contact Tracing (DCT) is a great example of information systems assisting societal problems. However, privacy concerns lead to reduced DCT adoption rates. When dealing with critical societal issues, policymakers seek to use various strategies, such as interventions and financial subsidies, to steer the behavors of individuals. This causes individuals to face a sophisticated decision-making process when coping with the public health crisis and the adoption of DCT, i.e., giving up privacy and freedom to gain information to remain healthy. In this paper, we consider a scenario, where policymakers allocate rewards to individuals to motivate their compliance with the interventions; And the individuals decide the optimal compliance effort based on their health state, privacy loss, interaction with their neighbors, and rewards. To tackle the trade-off between a number of individuals and policymakers, in this paper, we propose a Leader-Followers Mean-Field Game model to analyze this time-dependent, dynamic, and large-scale decision-making problem. The numerical results demonstrate that by allocating appropriate rewards, policymakers can play a role in guiding the behavior of individuals in various scenarios.

Due to the advantage of providing maximum data reliability at a given storage cost, $(k+r,k)$ maximum distance separable (MDS) array codes, in which all data symbols can be obtained by any $k$ of $k+r$ nodes, are extensively used in distributed storage systems. $(k+r,k,\alpha)$ piggybacking codes, which are a particular kind of $(k+r,k)$ MDS array codes, reaches low repair bandwidth for a failed single node and possesses small sub-packetization level $\alpha$. In this paper, by dividing the table formed by some instances of base MDS code into regions and carefully adding the data symbols in some regions as piggybacks to the parity symbols in some other regions, we design a new piggybacking code which has higher repair efficiency than the existing piggybacking codes in repairing each systematic node, where a node is said to be systematic if it keeps  portions of the raw message in reserve without encoding. The sub-packetization level is roughly equal to 2 times the number $r$ of parity nodes when the new piggybacking code attains the highest repair efficiency.

There has been a rapid advancement in developing computational models originating from the governing principles of quantum mechanics. Quantum computing approaches such as variational quantum algorithms (VQAs) and quantum-inspired algorithms have shown advantages across various applications. VQAs have been widely used as hybrid classical-quantum computational frameworks that are well-suited for NISQ advantages. Alternatively, the quantum-inspired evolutionary algorithm (QiEA) provides an efficient search space and parameter optimization solution, useful in handling problems that consist of large parameters. Following the recent trends, quantum-inspired technique, developments, limitations, and future scope are highlighted. Also, an approach towards advancing neuromorphic computation is discussed. Neuromorphic computation and systems are a fast developing domain inherited from fast and massively parallel processing, low energy consumption, high density of millions of neurons in a chip, and have found successful applications across spatio-temporal domain areas including brain data modelling. However, they require a fast and efficient parameter optimization process in large parameter space. Finally, future directions for the development of quantum-enhanced spiking neural networks and neuromorphic systems are outlined.

This paper describes a simulation-based approach for automated risk assessment of complex cyber-physical systems to support implementers of ISO 27005. The approach is based on systematic causeand-effect modelling of threats, their causes and effects, and the ways in which the effects of one threat can lead to other threats. In this way, the approach deals with inter-dependencies within the target system, automatically finding attack paths and secondary effect cascades, which generally are very complex and the source of many challenges for implementation of ISO 27005. The approach uses a knowledgebase describing classes of system assets and their possible relationships, along with the associated threats, causes and effects in a generic context. A target system can then be modelled in terms of related assets, describing the intended system structure and purpose (in the absence of any deviations). The knowledgebase is then used to identify which threats are relevant and create a cause-and-effect simulation of those threats. This allows threat likelihoods and risk levels to be found based on input concerning trust assumptions and the presence of controls in the system. The approach has been implemented by the open source Spyderisk project and validated by modelling a published case study of an attack on a steel mill. Given reasonable assumptions about security controls in place, the shortest, highest likelihood attack path found coincides with the published analysis. The case study demonstrates the strengths of the approach: transparency, reproducibility, and performance.

UTSOI MOS devices enable the scaling of the channel length of MOS Transistors through scaling of the channel thickness and oxide thickness. Additionally, as they have a planar architecture similar to the conventional Bulk MOSFETs, these UTSOI MOSFETs have a simpler fabrication process compared to 3-D Transistors like FinFETs. Also, their performance can be efficiently tuned by applying a bias through the BOX. This allows a significant dynamic modulation of the speed/power trade-off, which is a powerful knob at the circuit level.

The significance of system inertia, especially its nonuniform spatial distribution, is becoming paramount in the current power system scenario. This paper proposes a generalized two-level method to estimate the spatial inertia distribution, which is applicable for any given operating condition (or snapshot), as well as for ambient and transient power system conditions. The novelty of the proposed method lies in its mode-agnostic approach for estimating absolute-valued nodal inertias, which makes it more generalized than dominantmode-based methods, and more suitable for planning studies. Also, the proposed method does not require rate of change of frequency (ROCOF) measurements, or rate of change of power injections, which makes it immune to the noise associated in the estimation of these quantities. Additionally, the proposed method does not require defining near-zero ROCOF thresholds which is system-specific and a non-trivial problem. The applicability of the proposed method is tested through simulation studies performed in the IEEE-39 bus test system. The robustness of the method is also tested against modeling uncertainties and in the presence of measurement noise.

Opioid Use Disorder (OUD) is a major public health crisis in the US affecting over 2.4 million Americans across all age groups and backgrounds. Previous studies have developed machine learning (ML) and deep learning (DL) models with strong predictive performance for OUD; however, bias related to sociodemographic features remains unaddressed. In this study, we propose a bias mitigation algorithm, based on equality of odds (EO), to mitigate ML model bias in OUD prediction. Our algorithm adjusts the classification threshold to minimize the difference between specificity and recall across sociodemographic groups. We develop a neural network (NN) for OUD prediction and analyze bias by comparing the area under the curve (AUC) values. We implement our algorithm to mitigate bias related to gender, marital status, working condition, race, and income. The proposed algorithm achieves significant improvements in bias reduction: 16.96% for gender, 8.87% for marital status, 8.45% for working condition, 41.62% for race, and 0.20% for income. We also propose a fairness-aware weighted majority voting (WMV) classifier, achieving a recall of 92.68% and an accuracy of 90.21%. Furthermore, we evaluate the proposed methods using logistic regression and support vector machine classifiers with linear and radial basis function kernels, confirming the significant bias reduction and high performance of the WMV classifier for OUD prediction

A document by Mohammed Mahyoub . Click on the document to view its contents.