System level simulation of neuro-memristive circuits under variability are complex and follow a black-box neural network approach. In realistic hardware, they are often difficult to cross-check for accuracy and reproducible results. The accurate memristor model prediction becomes critical to decipher the overall the circuit function in wide range of non-ideal and practical conditions. In most neuro-memristive systems, crossbar configuration is essential for implementing multiply and accumulate calculations, that forms the primary unit for neural network implementations. Predicting the specific memristor model that best fits the crossbar simulations to make it explainable is an open challenge that is solved in this paper. As the size of the crossbar increases the cross-validation becomes an even more challenging. This paper proposes predicting the memristor device under test by automatically evaluating the I-V behavior using Random forest and Extreme Gradient Boosting algorithms. Starting with a single memristor model, the prediction approach is extended to memristor crossbarbased circuits explainable. The performance of both algorithms is analysed based on precision, recall, f1-score and support. The accuracy, macro average and weighted average of both algorithms at different operational frequencies are explored.

Electroencephalography (EEG) biometrics has garnered significant attention in recent years owing to its nonintrusive nature, real-time detection capabilities, concealment, and high complexity. Despite these promising attributes, the practical deployment of EEG-based identity recognition systems remains hindered by limited cross-day recognition performance. While some studies have reported cross-day recognition, they often suffer from slow recognition speeds, failing to meet the basic requirements for practical applications. To address this issue, we propose an unsupervised domain adaptation algorithm based on spatial pattern alignment for visual-evoked potential (VEP)-based identity recognition. This method employs rotational alignment of spatial patterns to correct cross-day spatial filters and utilizes forward selection to identify optimal sub-bands. By utilizing this approach, significant improvements of speed and accuracy in cross-day recognition can be achieved. We validate the proposed algorithm on three existing VEP datasets: Dataset I (25 subjects across 30 days), Dataset II (21 subjects across 5 days), and Dataset III (15 subjects across 200 days). The results demonstrate a significant superiority over the compared algorithms. Furthermore, we conduct online experiments with 15 individuals across over 1000 days, and the outcomes remain consistent. Analyzing the dataset over nearly three years in terms of temporal dimension, we observe evident template aging effect: 30 days > 200 days > 1000 days. However, the proposed method effectively mitigates template aging, resulting in minimal performance differences among the various datasets. The introduced algorithm substantially enhances speed and accuracy in cross-day recognition, paving the way for the long-term stability and practicality of online brainwave recognition systems.

Electric Vehicles (EVs) have already been playing a key role in reducing carbon emissions; and in the future, their importance is expected to grow even more. The number of EVs in developing countries is also increasing with time. Bangladesh, a developing country located in South Asia has witnessed a remarkable bloom in the number of Electric 3-wheelers vehicles. However, like other developing countries, Bangladesh faces great difficulties in balancing the net energy supply and demand. To address this issue of maintaining grid stability and fostering the ever-growing trend of EV adoption, this study introduces an optimization approach that regulates the total load of electric vehicles. Utilizing a heuristic method, this paper focuses on peak shaving and valley filling techniques for the specific EV charging patterns of Bangladesh. This approach incorporates three objective functions designed to amplify the difference between power generation and EV load during peak hours while reducing it during valley hours, all while considering the inclusion of weightage factors. Simulation data demonstrate that the implemented algorithm offers superior performance compared to the non-optimized system. Furthermore, we also estimated the electricity generation and EV load for the year 2030, and the result shows that the algorithm also function well with the forecasted data of electricity generation and demand.

This study examines the effects of macroeconomic policies on financial markets using a novel approach that combines Machine Learning (ML) techniques and causal inference. It focuses on the effect of interest rate changes made by the US Federal Reserve System (FRS) on the returns of fixed income and equity funds between January 1986 and December 2021. The analysis makes a distinction between actively and passively managed funds, hypothesizing that the latter are less susceptible to changes in interest rates. The study contrasts gradient boosting and linear regression models using the Double Machine Learning (DML) framework, which supports a variety of statistical learning techniques. Results indicate that gradient boosting is a useful tool for predicting fund returns; for example, a 1% increase in interest rates causes an actively managed fund's return to decrease by-11.97%. This understanding of the relationship between interest rates and fund performance provides opportunities for additional research and insightful, data-driven advice for fund managers and investors.

This paper introduces a dataset capturing brain signals generated by the recognition of 100 Malayalam words, accompanied by their English translations. The dataset encompasses recordings acquired from both vocal and sub-vocal modalities for the Malayalam vocabulary. For the English equivalents, solely vocal signals were collected. This dataset is created to help Malayalam speaking patients with neuro-degenerative diseases. This dataset not only contributes to the advancement of braincomputer interface technology but also holds promise in fostering effective communication solutions for individuals with restricted verbal abilities.

Optical polarizers are essential components for the selection and manipulation of light polarization states in optical systems. Over the past decade, the rapid advancement of photonic technologies and devices has led to the development of a range of novel optical polarizers, opening avenues for many breakthroughs and expanding applications across diverse fields. Particularly, two-dimensional (2D) materials, known for their atomic thin film structures and unique optical properties, have become attractive for implementing optical polarizers with high performance and new features that were not achievable before. This paper reviews recent progress in 2D-material-based optical polarizers. First, an overview of key properties of various 2D materials for realizing optical polarizers is provided. Next, the state-of-the-art optical polarizers based on 2D materials, which are categorized into spatial-light devices, fiber devices, and integrated waveguide devices, are reviewed and compared. Finally, we discuss the current challenges of this field as well as the exciting opportunities for future technological advances.

We observe that γ-ray radiation affects the formation of the conducting bridge in Ag/Ti/Al:HfO2/Pt devices. We suggest that the γ-ray breaks Hf-O bonds and affects the properties of metal/insulator interfaces. The radiation-induced interfacial layers promote the transition from write-once-read-many times (WORM) to reversible switching memories. The devices that undergo a higher radiation exposure exhibit a higher forming voltage that we could exploit to sense radiation; an electrical circuit to harness this phenomenon is also proposed. We also observe that the devices exhibit self-healing behavior, where the forming behavior restores once the radiation energy is released. The switching mechanism is explained and proposed to elucidate this phenomenon. This study not only provides insight into the development of memristor devices for space application but also their potential as multipurpose elements for reconfigurable circuits.

To estimate the electrical dynamic characteristics and control coefficients of distributed energy resources (DERs) in microgrids, we introduce an improved physics-informed neural network (PINN) for parameters estimation of microgrid devices. The novelty of our approach lies in two key advancements: First, we introduce a data transformation methodology that significantly accelerates the training process, achieving an increase in speed of up to 82.87% compared to the original PINN. Second, we establish ordinary differential equations (ODEs) frameworks that are adaptable to both inverter-based and synchronous-based DERs, thereby enhancing the generalizability of the proposed method. These advancements enhance the application of PINN in microgrids, paving the way for instructive simulation modeling. For analysis, we established typical microgrid systems using real-time digital simulation (RTDS) to generate the necessary data. This paper explores the effectiveness of PINNs across varying ODEs, showcasing the algorithm's adaptability to diverse mathematical models. Additionally, we illustrate the accuracy of PINN in parameter estimation for both synchronous and inverterbased DERs, with our enhanced approach achieving parameter deductions with less than a 5% error margin in a training session.

The advent of flexible, compact, energy-efficient, robust, and user-friendly wearables has significantly impacted the market growth, with an estimated value of 61.30 billion USD in 2022. These wearable sensors have revolutionized in-home health monitoring by warranting continuous measurements of vital parameters. Ultrasound is used to non-invasively, safely, and continuously record vital parameters. The next generation of smart ultrasonic devices for healthcare integrates microelectronics with flexible, stretchable patches and bodyconformable devices. They offer not only wearability, and user comfort, but also provide higher tracking accuracy of immediate changes of cardiovascular parameters. Moreover, due to the fixed adhesion to the skin, errors derived from probe placement or patient movement are mitigated, even though placement at the correct anatomical location is still critical and requires a user's skill and knowledge. In this review, the steps required to bring wearable ultrasonic systems into the medical market (technologies, device development, signal-processing, in-lab validation, and, finally, clinical validation) are discussed. The potential of the next generation of vascular ultrasound and its future research directions offer many possibilities for modernizing vascular health assessment and the quality of personalized care for home and clinical monitoring.

Human pose estimation has a variety of applications in action recognition, human-robot interaction, motion capture, augmented reality, sports analytics, and healthcare. There is a substantial stream of datasets and deep learning-based models to attain robust human pose estimation within the visible domain. Nonetheless, there are certain obstacles in this domain, including insufficient illumination and privacy concerns. These issues can be addressed using thermal cameras. However, only a limited number of annotated thermal human pose datasets are available to train data-hungry deep learning models. In this regard, we introduce a novel open-source thermal human pose dataset named OpenThermalPose. The dataset contains 6,090 thermal images and 14,315 annotated human instances. The annotations include bounding boxes and 17 anatomical keypoints, following the annotation format of the MS COCO dataset. The dataset covers various fitness exercises, multiple-person activities, and outdoor walking in different locations and weather conditions. As a baseline, we trained and evaluated YOLOv8-pose models on our dataset. We have made the dataset, source code, and pre-trained models publicly available at https://github.com/IS2AI/OpenThermalPose to bolster research in this area.

The growing relevance of multi-agent systems has drawn increasing focus on communication-efficient filters for collaborative target tracking to alleviate the system's communication burden. While the event-triggered (ET) mechanism can improve communication efficiency in collaborative tracking, an inevitable trade-off exists between estimation accuracy and communication cost in ET filters. This paper proposes a fast and accurate ET diffusion-based filter for real-time multi-agent collaborative tracking, aiming to reduce the system's data transmission without notable compromise in tracking performance. The proposed filter achieves improved tracking accuracy, reduced data transmission, and accelerated convergence using an error-minimized ET cubature information filter (CIF) for local estimation, and a correlation-aware diffusion strategy for global fusion. Our algorithm's stability has been proved. Results show that our algorithm can reduce estimation error by 21.6-72.18%, decrease computation time by 47.2-62.55%, and improve communication efficiency by 34.42-49.65% compared to mainstream collaborative tracking approaches.

The cell-to-cell coupling in a reconfigurable intelligent surface (RIS) is very different from a periodic structure, where coupling effects can be precisely evaluated via full-wave analysis with periodic boundary conditions. We propose a novel method based on convolutional neural networks (CNNs), to predict the contribution of mutual coupling on the near-zone tangential electric field of every RIS unit cell that characterizes its scattering. Our CNN model incorporates an attention mechanism based on the squeeze-and-excitation block module, enhancing its capability to discern and quantify coupling effects, especially from neighboring cells surrounding the unit cell of interest. The predictions of the model enable the computation of RIS scattered fields, fully accounting for the aperiodic nature of an RIS. Comparisons to finite-element analysis confirm that our computed fields are accurate at any point and for any RIS configuration. Furthermore, our trained model can be retrained through transfer learning, to accurately and efficiently predict cell-to-cell coupling under different incident wave conditions, utilizing only a reduced number of training data. Therefore, the proposed method is a valuable tool for various practical applications, such as synthesizing RIS scattered field patterns and evaluating the performance of RIS-enabled channels.

Memristors with various switching characteristics could realise low-power reconfigurable electronics. This paper reports the coexistence of nonvolatile and volatile switching characteristics in Ag/Ti/ZnO/TiN CBRAM memristor devices and examines the feasibility of the cycle-to-cycle fluctuation of both characteristics for memory and random number generator (RNG) applications. A circuit design exploiting the nonuniformity to produce random bits is proposed. We show that random bits can be produced by the circuit regardless of the nonvolatile switching characteristics; nevertheless, volatile behavior may be preferred for low-power operation. Nonvolatile-tovolatile switching transformation can be achieved by inserting an excessively thick barrier layer or applying a low current compliance to form a metastable conducting bridge. We found a close correlation between the statistical distribution of the switching nonuniformity and the randomness parameter for making reliable random bits.

The implementation of two-terminal memristors could decrease the dependency on the transistors in logic circuits, enabling high-density circuits and reducing power consumption due to the memristor's smaller feature size and non-volatility, respectively. We propose universal logic function designs using ZnO-based transparent memristors. A highly accurate fitting of set and reset curves is achieved by employing Verilog model and used for circuit-level simulations. OR, NOR, NAND, and AND gates can be designed by simply tuning the memristor and pullup resistor while using a similar circuit configuration. Based on this facile configuration, more complex circuits, such as XOR and XNOR gates as well as half-adder and half-subtractor combinational logic circuits can be achieved. This work highlights a promising alternative for reconfigurable logical integrated circuits.

This paper presents a novel Ground Penetrating Radar (GPR)-based framework for precisely estimating rootzone soil moisture, a key parameter in precision agriculture. The approach is structured as follows: First, we generate a synthetic dataset using gprMax, carefully calibrated against realworld data to reflect actual soil conditions. Feature engineering techniques are then employed to extract meaningful features from the GPR signals, followed by a rigorous selection process to identify the most effective machine learning (ML) model for soil moisture prediction. The proposed framework is crucial for optimizing irrigation practices and water resource management, directly contributing to more sustainable agriculture. Finally, we validate our model by integrating synthetic data with real GPR data collected at the SoilX Lab at Worcester Polytechnic Institute (WPI), enhancing prediction accuracy and generalization capability. By ensuring our model's applicability to diverse agricultural environments, we aim to support efficient farming practices that lead to better crop yields and food security.

The feasibility of memristor devices as a controllable resistor element for making tunable analogue filters is investigated by employing the Stanford Resistive Random Access Memory (RRAM) memristor model. Replacing the resistor in a conventional filter with a memristor allows the filter to have a variable range, this would solve the limitation where the conventional analogue filters have only a single filter range due to the fixed values of their constituent elements. This paper reports the results of fitting the Al: HfO2 memristor using the Stanford RRAM model, and the fitted results can be perfectly matched to real experimental results. Memristor-based filters can obtain at least two different filter ranges with the same filtering effect as conventional filters.

This paper explores EfficientNet smaller models for a multiclassification task of corrosion types in above-ground storage tanks through transfer learning and finetuning approaches. Data augmentation was used to increment the data an oil and gas company provided, reaching a dataset of around 5000 images. The images were stored in Google Drive and imported by Colab to obtain the models using TensorFlow and Keras. After the hyperparameters’ tuning a transfer learning model was selected and explored with fine tuning. The EfficientNetB0 model delivered from fine-tuning accomplished 94% performance. This work is the first attempt to deploy an artificial vision automatic tool for being implemented during tank inspection in the industrial sector. In a further development, this model can be coupled with one based on object detection for the remote identification of failures due to external corrosion during tank inspection; improving safety and reliability in the oil and gas industry.

Next-generation cellular technologies, commonly referred to as the sixth generation (6G), are envisioned to support a higher system capacity, better performance, and network sensing capabilities. The terahertz (THz) band is one potential enabler to this end due to the large unused frequency bands and the high spatial resolution enabled by the short signal wavelength and large bandwidth. Different from earlier surveys, this paper presents a comprehensive treatment and technology survey on THz communications and sensing in terms of advantages,