Weather forecasting is essential for agriculture, disaster management, energy planning, and infrastructure protection. Using a 96k-record hourly dataset targeting six variables (temperature, humidity, wind speed, wind bearing, visibility, pressure) traditional numerical weather prediction (NWP) systems demand high-performance computing resources and long-term historical data, making them impractical for newly deployed weather stations with limited historical data availability. To address this limitation, this research proposes and evaluates deep learning-based architectures for weather time series forecasting under varying data availability conditions. Specifically, three standalone models and several novel hybrid models are tested across four temporal scales: 10 years, 5 years, 1 year, and 1 month. The study shows that hybrid models consistently outperform standalone models in terms of accuracy and robustness. Among them, the Temporal Convolutional Network Long Short-Term Memory (TCN-LSTM) hybrid achieves the best performance, with an$R^{2}$of 0.88 for short-term and 0.99 for longterm temperature forecasting, while remaining computationally efficient. These models offer scalable, accurate forecasting solutions well-suited for early deployment in resource-constrained and data-scarce environments.

The primary objective of this work is to develop the charged-type memristor emulator for a higher frequency of operation. Here, an active current mode block DVCCTA (Differential Voltage Current Conveyor Transconductance Amplifier) is utilized along with two resistors and one capacitor to build a floating charged type memristor emulator. A straightforward switch can enable circuit operation in both Incremental and Decremental modes. The resilience of the suggested circuit is confirmed through Monte Carlo simulations to ensure its stability and performance under varying conditions. PSPICE simulation is executed for diverse frequencies and capacitors using 180nm TSMC technology. The simulation outcome supports the conclusions drawn from this literature's conceptual and frequency response analysis. To study the memory attributes of a memristor, a retention test has been carried out for both incremental and decremental modes. Finally, the proposed design is employed in the Schmitt trigger circuit to check its performance.

In the recent advancements of 3D integrated circuits, the adoption of tapered-shaped through silicon via (TSV)- bumps emerges as an appealing substitute for conventional cylindrical structure owing to their reduced parasitic effects and space-efficient structure. The copper and carbon nanotube (Cu-CNT) composite based tapered TSV-bump structure offers superior electrical conductivity, thermal stability, and mechanical strength compared to standalone Cu or CNT materials, making them an ideal choice. It is crucial to model the electrical characteristics of Cu-CNT based TSVs-bump, and this can be achieved by employing the effective complex conductivity approach. A comprehensive model, denoted by resistance-inductance-conductance-capacitance (RLGC), is implemented to consider the influences of both the inter-metal dielectric and bump effects. The computation of TSV-bump parasitic is accomplished by applying the current continuity expression, utilizing the partial inductance method, and segmenting infinitesimally thin slices of the bump within a triangular configuration of tube assemblies. In order to validate the proposed model, the quantitative values of TSV-bump parasitics are compared with the EM simulations, revealing a notable conformity with an average deviation of only 2.9%. Further, the crosstalk-induced delay, insertion loss (S21), and reflection loss (S11) for the Cu, CNT, and composite Cu-CNT based tapered TSV-bump structure have been investigated by using a driver-via-load (DVL) in 7nm technology. Furthermore, when compared to the Cu and CNT material, the Cu-CNT composite based tapered TSV-bump depicts an average improvement of crosstalk, insertion, and reflection losses by 39.12%, 48.23% and 57.19%, respectively.

Modulation is a fundamental technique for transmitting signals between a transmitter and receiver in wired and wireless scenarios. In a non-cooperative communication scenario, the receiver should be capable of classifying the modulation type efficiently. Automatic Modulation Classification (AMC) is designed to monitor the radio-frequency spectrum and for making effective transmission decisions. Earlier AMC methods relied on feature extraction and likelihood-based approaches. However, with the rise of Deep Learning (DL), AMC has gained momentum due to its superior predictive capabilities. In this work, we propose a custom artificial neural network (ANN) for modulation classification. We analyze the impact of a deep learning model on AMC and how different modulation techniques influence its accuracy.

In this paper, we consider a hybrid satellite-terrestrial network (HSTN) where a multiantenna satellite communicates with a ground user equipment (UE) with the help of multiple amplify-and-forward (AF) three-dimensional (3D) mobile unmanned aerial vehicle (UAV) relays. Herein, we employ a stochastic mixed mobility (MM) model to deploy mobile UAV relays in a 3D cylindrical cell with UE at its ground centre. Taking into account the multiantenna satellite links and the random 3D distances between UAV relays and UE, we analyze the outage probability (OP) of considered system under an opportunistic UAV relay selection policy. We further carry out asymptotic OP analysis to present insights on system diversity order. Moreover, we compare the performance of proposed 3D mobile UAV relaying with the fixed altitude mobile UAV relaying as well as the fixed distance static relaying schemes. The analysis will be verified through simulations.

This work aims to enhance the quality of narrowband (0–4 kHz) voice signal in terms of frequency components, i.e. missing high‐frequency components in a range of 4–8 kHz. The proposed artificial bandwidth extension framework uses the optimisation. In this context, a signal model is used to get a better representation of wideband (0–8 kHz) information of a signal. The optimisation is used to obtain the synthesis filter for a given signal model, which is used to synthesise the high‐band (4–8 kHz) signal. The discrete Fourier transform addition is performed to add the narrowband signal and estimated high‐band signal for removing the leaked information from the synthesis filter and non‐ideal low pass filter. Gain adjustment is performed on the estimated high‐band signal to make its energy equal to the true high‐band signal. Non‐stationary characteristics of speech signals generate an assorted variety in synthesis filters and corresponding gain. For this, a deep neural network (DNN) is used to estimate the synthesis filter and gain by using the given narrowband information. The authors analyse the performances of the DNN model on two data sets. Objective and subjective analyses are carried out on these data sets.