Recently myokinetic interfaces have been proposed to exploit magnet tracking for controlling bionic prostheses. This interface derives information about muscle contractions from permanent magnets implanted into the amputee's forearm muscles. Machine learning models have been mapped on Field Programmable Gate Arrays (FPGAs) to track a single magnet, achieving good precision and computational efficiency, but consuming a large area and hardware resources. To track several magnets, here we propose a novel solution based on dynamic partial reconfiguration, switching three prediction models: a linear regressor, a radial basis function neural network, and a multi-layer perceptron neural network. A system with five magnets and 128 magnetic sensor inputs was used and experimental data were collected to train a system with five hardware predictors. To reduce the complexity of the models, we applied principal component analysis, ranking by correlation the number of inputs of each model. This run-time reconfigurable solution allows the circuits to be reconfigured in order to select the most reliable predictor model for each magnet while the rest of the circuit continues to operate extracting the most significant information from the captured signals. Thus, the proposed solution remarkably reduces the hardware occupation and improves the computational efficiency compared to previous solutions.

This paper proposes a direct data interpolation method for predicting ferrite core loss in power electronics. The core idea is to accurately construct the B-H loop shape given arbitrary Hfield waveforms. Unlike conventional loss models like the Steinmetz equation and its variations, or recent neural network methods like MagNet that leverage periodic B-H waveforms and frequencies, this method employs a data segmentation and reconstruction process to fully exploit the distinctive features of the B-H loop under varying conditions. The source data is from Princeton University's MagNet open-source datasets.

This paper describes integration of a software defined antenna (SDA) and a software defined radio (SDR) to enable dynamic and automatic selection of physical propagation channels. The SDA, an agile aperture antenna (A3), provides monolithic microdiversity, i.e. multiple antenna states within a single structure and within a space having maximum dimension of approximately one wavelength. In this way physical channel selection occurs outside, and therefore augmenting, the front-end electronics. The independent controllable parameters for channel access include: frequency, polarization (full Poincaré sphere), and pattern (in two dimensions). Several different tests were conducted to maximize a quality figure of merit calculated by the radio. Both library-based and evolutionary search techniques were used, resulting in short term channel improvements on the order of 10 dB, and long term fully optimized improvements on the order of 20 dB.

Recent findings reveal that a simple one-layer linear model can significantly outperform transformer-based models. To further enhance time series forecasting performance, we introduce a feedforward neural network (FNN) -based deep learning model, which surpassed most existing models. The novel FNN-based model presents three innovations: input and output structure, FNN division and combination, and an FNN sharing mechanism. The model utilizes three inputs: historical values, historical timestamps, and future timestamps, in order to predict the future demand as its output. FNN division means different inputs corresponding to different FNNs, and the outputs of these FNNs are combined to form the final prediction. The FNN sharing mechanism involves mapping each column (or row) of the matrix inputs to the column (or row) using a common FNN, termed FNN kernel. This FNN neural network framework reduces computational complexity and the number of weights significantly while still providing interpretability and strong performance. Our comparative experiments demonstrate that all three models outperform existing models on two real-world datasets.

A document by Daniel Chaver. Click on the document to view its contents.

An effective and important technique, Deep Learning has been found to be recurrently used in different fields, including computer vision, machine vision, and natural language processing. As a result of this technology, deepfakes generate visual media of a particular person that are nearly similar to the original material. Over the past few years, a variety of studies have examined how deepfakes operate, and numerous methods utilizing deep learning have come forward to detect both deepfakes videos and images. In this work, the dangers deepfakes present to corporate cybersecurity are investigated, underscoring the need to understand the technology related to deepfakes and the importance of constructing effective detection and mitigation methods. The article includes coverage of the merits and shortcomings of several methods for locating deepfakes, including deepfake detection algorithms along with human review and contextual analysis. A technique that unites automated detection with human judgment is critical for discovering deepfakes. The article brings to attention the necessity for businesses and individuals to strengthen their awareness and evasion capabilities against deepfake deception, decrease the risk that deepfakes can get, and take advantage of the potential opportunities deepfakes might create. Key takeaways reveal that facial landmark analysis, lip synchronization detection, and micro-expression analysis play key roles in deepfake detection and that a zero-trust and multi-factor authentication protocol is necessary to address the risks associated with deepfakes.

This article focuses on analyzing SplitFed Learning against model poisoning vulnerability and developing methods to protect such a system against these attacks. SplitFed learning is a distributed learning paradigm where a neural network model is split between clients and the server, contrasting with traditional Federated Learning. SplitFed learning enables enhanced security and privacy of data, and clients do not need to perform heavy computation in model training, as they only need to train a part of the model. This approach ensures that the model can make precise predictions while maintaining the confidentiality of sensitive information. In addition to implementing a SplitFed Model, the paper proposes a distance-based method that can poison SplitFed Learning-based systems. Subsequently, this paper develops a novel prevention strategy based on robust statistical properties of the sample. To test the proposed methodology, as a test case, we have employed the image cell dataset of the malaria parasite. By addressing the impacts of adversarial attacks, this paper contributes to the advancement of deep learning techniques.

In this work, we tackle the estimation of elevation and azimuth direction of arrival (DOA) of uncorrelated wideband (WB) and statistically non-stationary (NS) sources located in the far-field of a sensor array in both free-field and multipath propagation environments. For the case of free-field propagation, we derive the closed-form expression of the Cramer-Rao Bound (CRB) that is applicable to both the overdetermined and underdetermined cases of DOA estimation. We also establish the condition for the existence of the CRB along with a discussion on the dependence of CRB on the signal-to-noise ratio and the upper bound on the total number of WB and NS sources that can be resolved theoretically. Further, we develop a coarray-based DOA estimation algorithm that uses the enhanced degrees of freedom (DOF) of the difference coarray of a non-uniform planar array to localize more uncorrelated WB and NS sources than the number of physical sensors. In the multipath scenario, the correlation between the direct signal and the reflected signal is tackled by proposing a new focusing method based on the decomposition of the array steering function using 2D Fourier basis functions. The proposed focusing approach can be applied to any realworld arbitrary planar array and does not require the initial estimate of the DOAs of the WB and NS sources. After focusing, frequency smoothing is performed to decorrelate the direct and reflected signals. After decorrelation, we propose two coarraybased techniques to identify the time-frequency bins that are dominated by the direct signals which are then used to estimate the DOA of the sources emitting the direct signals. Numerical examples verify the effectiveness of the proposed methods.

In addition to the growing need for wireless connectivity, the advancement of 6G technology is also driven by the dynamic and varied requirements of different vertical sectors. The vision of 6G extends beyond enabling pervasive and ubiquitous wireless intelligence and addresses seamless communication between humans, machines, and everything in between. This paper introduces key performance indicators (KPIs) as crucial metrics for identifying vertical demand and conducting digitalization requirements analysis within the energy sector. Moreover, it highlights the smart grid use cases with their specified communication KPI values. Finally, the digitization of energy systems will be discussed, along with its potential challenges, opportunities, and solutions.

A document by Luis Hernandez Corporales . Click on the document to view its contents.

Real World Assets (RWAs) have emerged as a pivotal concept in the transformation of global financial markets, driven by blockchain technology and the process of tokenization. By digitizing physical assets and traditional financial instruments, tokenization enhances liquidity, accessibility, and efficiency in financial transactions. Escrow mechanisms, particularly those enabled by blockchain-based smart contracts, play a crucial role in ensuring the secure and transparent transfer and management of these tokenized assets. This article examines the connection between RWAs and escrow mechanisms, exploring foundational concepts, types, and advantages of tokenizing RWAs, as well as the functions and innovations of escrow mechanisms in this context. The paper analyzes the synergies between tokenization and escrow services, discusses practical applications and case studies, and addresses the comprehensive challenges and risks involved. Finally, it offers future perspectives on technological advancements, regulatory developments, market growth, and collaborative ecosystem development in the integration of RWAs and escrow mechanisms.

Electric grid operators are adept at handling complexity and uncertainty. However, with increasing introduction of renewable generation, distributed energy resources, and more frequent severe weather events, operators will experience new workload and challenging decision scenarios. This paper quantifies risks and benefits from an operator’s perspective of introducing weather based forecast Dynamic Line Ratings (DLR) using variable wind conditions in addition to ambient temperature to relieve transmission congestion and facilitating more offshore wind (OSW). A concept of operations (CONOPS) applied to a forecast DLR implementation and its integration with OSW is defined. A method for evaluating tradeoffs of derating to make the rating more conservative but decreasing the benefit was developed and applied to a case study for two existing overhead transmission lines on Long Island, New York. The CONOPS uses historical day-ahead and hour-ahead High Resolution Rapid Refresh weather forecasts and weather station data to support planning and real-time operations. The analysis determines the risk of downgrades in real-time operational rating compared to the forecast and quantifies the frequency and severity of lastminute downgrades. The risk is compared against the benefits in increased capacity to provide insights on the additional amount of uncertainty DLR and OSW will add to the operator’s workload.

A document by Anushri Pal. Click on the document to view its contents.

We address the problem of characterising the aggregate flexibility in populations of electric vehicles (EVs) with uncertain charging requirements. Building on previous results that provide exact characterisations of the aggregate flexibility in populations with known charging requirements, in this paper we extend the aggregation methods so that charging requirements are assumed to be uncertain, but sampled from a known distribution. In doing so we, construct distributionally robust aggregate flexibility sets, sets of aggregate charging profiles, over which we can provide probabilistic guarantees, that sampled populations will be able to track. By leveraging certain measure concentration results that establish powerful finite sample guarantees, we are able to give tight bounds on these robust flexibility sets, even in low sample regimes that are well suited for aggregating small populations of EVs. We detail explicit methods of calculating these sets and provide numerical results that validate the theory developed here.

Wearable technologies have been demonstrated to have value in the objective assessment of Parkinson's disease. However, certain symptoms such as freezing of gait are challenging to monitor using current approaches. Data augmentation, wherein synthetic or simulated data is added to real world training sets to increase their size and diversity, has emerged as an approach to bolster the accuracy of machine learning models in other healthcare domains such as medical imaging. In this paper, we evaluate the use of generative adversarial networks to create synthetic accelerometry data representing both typical and freezing gait patterns. Our preliminary results suggest that these synthetic datasets can mimic realistic movement patterns of individuals with Parkinson's disease. In the future, we will assess the impact that integrating these synthetic data into training sets has on the accuracy of freezing of gait detection algorithms.

This paper explores the potential use of ultra-wideband (UWB) wireless sensors in detecting the transition of food items from a frozen state to thawed state (or vice-versa), marking a significant advancement in monitoring the thawing process. By exploiting the drastic change in complex permittivity at microwave frequencies (we use frequencies close to 4GHz) during the solid-liquid state transition, this paper introduces a novel approach in ensuring frozen food safety, and in enhancing the efficiency of the frozen food industry and cold chain transportation. The developed system, capable of operating through food packaging, offers a non-invasive, cost-effective solution for real-time monitoring, addressing the limitations of conventional temperature-based and timing-based methods widely used today in household and professional cooking and in the food industry. Our findings from the raw UWB channel impulse responses (CIR) and computed similarity scores indeed show significant promise and validate the feasibility of the proposed system with various realworld applications.

Fourteen students and the professors attended the "Accounting in integrated management systems" class using Meta Quest 2 VR headsets during seven 1-hour sessions, one session per week. Different aspects were analyzed during the sessions, including audiovisual quality, comfort, sickness, immersion, presence, fatigue, cognitive load, and useful of the technology for the academic purposes.

Conventional attention mechanisms, particularly those applied in Transformer architectures, have achieved remarkable success across various domains, including natural language processing, computer vision, and multimodal data integration. However, these methods often focus primarily on intra-cluster relationships, limiting their capacity to effectively capture inter-cluster dependencies that are essential in data with complex, heterogeneous structures. To address this gap, we propose a novel approach utilizing energy-based attention mechanisms to enhance inter-cluster interactions. Our study introduces and evaluates several energy well configurations, including Gaussian, Lorentzian, and softmax exponential models, which are tailored to adjust attentional focus dynamically across clusters. This work demonstrates that energy-based inter-cluster attention mechanisms not only improve interpretability but also outperform traditional attention in both accuracy and computational efficiency. Experimental results on text and image datasets confirm the efficacy of our approach, making it a promising alternative to standard attention mechanisms in neural networks.