Pretraining has become a standard technique in computer vision and natural language processing, which usually helps to improve performance substantially. Previously, the most dominant pretraining method is transfer learning (TL), which uses labeled data to learn a good representation network. Recently, a new pretraining approach – self-supervised learning (SSL) – has demonstrated promising results on a wide range of applications. SSL does not require annotated labels. It is purely conducted on input data by solving auxiliary tasks defined on the input data examples. The current reported results show that in certain applications, SSL outperforms TL and the other way around in other applications. There has not been a clear understanding on what properties of data and tasks render one approach outperforms the other. Without an informed guideline, ML researchers have to try both methods to find out which one is better empirically. It is usually time-consuming to do so. In this work, we aim to address this problem. We perform a comprehensive comparative study between SSL and TL regarding which one works better under different properties of data and tasks, including domain difference between source and target tasks, the amount of pretraining data, class imbalance in source data, and usage of target data for additional pretraining, etc. The insights distilled from our comparative studies can help ML researchers decide which method to use based on the properties of their applications.

We propose and develop comprehensive foundations for the analysis of nonlocal radiating systems using a special momentum-space approach. Part I focuses on the rigorous mathematical basis of the theory and their conceptual implications.

We propose and develop comprehensive foundations for the analysis of nonlocal radiating systems using a special momentum-space approach. Part II focuses on building some basic examples and applications of the theory explicated in Part I.

Postprint accepted on 21 May 2020 for publication in IEEE 92nd Vehicular Technology Conference. (c)2020 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

Neural architecture search, which aims to automatically search for architectures (e.g., convolution, max pooling) of neural networks that maximize validation performance, has achieved remarkable progress recently. In many application scenarios, several parties would like to collaboratively search for a shared neural architecture by leveraging data from all parties. However, due to privacy concerns, no party wants its data to be seen by other parties. To address this problem, we propose federated neural architecture search (FNAS), where different parties collectively search for a differentiable architecture by exchanging gradients of architecture variables without exposing their data to other parties. To further preserve privacy, we study differentially-private FNAS (DP-FNAS), which adds random noise to the gradients of architecture variables. We provide theoretical guarantees of DP-FNAS in achieving differential privacy. Experiments show that DP-FNAS can search highly-performant neural architectures while protecting the privacy of individual parties. The code is available at https://github.com/UCSD-AI4H/DP-FNAS

This paper introduces a systematic design method for decoupling elements, which can significantly improve the isolation between two co-located antennas, e.g. between transmit and receive antennas of an in-band full-duplex system. The design method applies the theory of characteristic modes for controlling the phase and amplitude of the scattered fields of the decoupling element, in order to optimally cancel the original incident fields which couple to the receiving antenna. We describe concisely the effects that characteristic angle, modal near-field, and modal excitation of the decoupling element have on the antenna isolation. For validating the proposed method, a planar wavetrap is designed and the isolation improvement verified with full-wave simulations. When we use the proposed method to optimize a wavetrap that is placed between two co-located patch antennas, we obtain an improvement of the isolation between the antennas by 33 dB at the centre frequency of their operational frequency band, and at least 12-dB improvement across the whole 142-MHz operational bandwidth of the two antennas. As a benchmark, the wavetrap is replaced by an absorber occupying 10 times the volume of the wavetrap. The absorber gives only 6 dB of isolation improvement, substantiating the effectiveness of the proposed wavetrap method.

Research in quantum computing hardware has shown that quantum bits can be implemented with superconducting circuits, these qubits could be used in other applications such as quantum devices for electronic systems. In this paper a nonlinear parameterised two qubits quantum device is presented, the current - voltage characteristic of this element is similar to an adjustable memristor when its parameters are changed.

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

Circuit models for multiphase coupled inductors are summarized, compared, and unified. Multiwinding magnetic structures are classified into parallel-coupled structures and series-coupled structures. For parallel-coupled structures used for multiphase inductors, the relationships between a) inductance-matrix models, b) extended cantilever models, c) magnetic-circuit models, d) multiwinding transformer models, e) gyrator-capacitor models, and f) inductance-dual models are examined and discussed. These models represent identical physical relationships in the multiphase coupled inductors, but emphasize different physical aspects and offer distinct design insights. The circuit duality between the series-coupled structure and the parallel-coupled structure is explored. Design equations for interleaved multiphase buck converters based on these models are streamlined and summarized, and a simplified equation showing the relationships between current ripple with and without coupling is presented. The models and design equations are verified through theoretical derivation, SPICE simulation, and experimental measurements.

The rapid rise of cyberattacks and the gradual failure of traditional defense systems and approaches led to using Machine Learning (ML) techniques to build more efficient and reliable Intrusion Detection Systems (IDSs). However, the advent of larger IDS datasets has negatively impacted the performance and computational complexity of ML-based IDSs. Many researchers used data preprocessing techniques such as feature selection and normalization to overcome such issues. While most of these researchers reported the success of these preprocessing techniques on a shallow level, very few studies have been performed on their effects on a wider scale. Furthermore, the performance of an IDS model is subject to not only the utilized preprocessing techniques but also the dataset and the ML algorithm used, which most of the existing studies give little emphasis on. Thus, this study provides an in-depth analysis of feature selection and normalization effects on various IDS models built using two IDS datasets namely, NSL-KDD and UNSW-NB15, and five different ML algorithms. The algorithms are support vector machine, k-nearest neighbor, random forest, naive bayes, and artificial neural network. For feature selection and normalization, the decision tree wrapper-based approach, which tends to give superior model performance, and min-max normalization methods were respectively used. A total of 30 unique IDS models were implemented using the full and feature-selected copy of the datasets. The models were evaluated using popular evaluation metrics in IDS modeling, intra- and inter-model comparisons were performed between models and with state-of-the-art works. Random forest achieved the best performance on both NSL-KDD and UNSW-NB15 datasets with prediction accuracies of 99.87% and 98.5%, as well as detection rates of 99.79% and 99.17% respectively, it also achieved an excellent performance in comparison with the recent works. The results show that both normalization and feature selection positively affect IDS modeling with normalization shown to be more important than feature selection in improving performance and computational time. The study also found that the UNSW-NB15 dataset is more complex and more suitable for building and evaluating modern-day IDS than NSL-KDD.

This paper reports the fundamentals and SPICE implementation of the dynamic memdiode model (DMM) for the conduction characteristics of bipolar resistive switching devices. The model equations are implemented in the LTSpice simulator using an equivalent circuital approach with behavioral components and sources.

This work analyses the different phases of Covid-19 outbreak in India and performs progress of the disease spread. The data is collected from John Hopkins epidemiological data providing the latest data from January 31,2020 to May 31, 2020. A simple mathematical model is developed to gather a quantitative picture of the epidemic spreading with limited reference data. Further, we performed an analysis and forecast of the disease spread in different phases of lock down in the country. The profound model predictions considering the overall data exhibit that the numbers to reach peak between 28 August 2020 to 6 Sep 2020. As the pandemic still increases the number of infected cases, different quarantine levels would serve as an effective measure in containing the spread much earlier than the other similar cases.

Artificial neural networks (ANNs) have won numerous contests in pattern recognition, machine learning, and artificial intelligence in recent years. The neuron of ANNs was designed by the stereotypical knowledge of biological neurons 70 years ago. Artificial Neuron is expressed as f(wx+b) or f(WX). This design does not consider dendrites’ information processing capacity. However, some recent studies show that biological dendrites participate in the pre-calculation of input data. Concretely, biological dendrites play a role in extracting the interaction information among inputs (features). Therefore, it may be time to improve the neuron of ANNs. In this study, some dendritic modules with excellent properties are proposed and added to artificial neurons to form new neurons named Gang neurons. E.g., The dendrite function can be expressed as Wi,i-1Ai-1 ○ A0|1|2|…|i-1 . The generalized new neuron can be expressed as f(W(Wi,i-1Ai-1 ○ A0|1|2|…|i-1)).The simplified new neuron be expressed as f(∑(WA ○ X)). After improving the neurons, many networks can be tried. This paper shows some basic architecture for reference in the future. Up to now, others and the author have applied Gang neurons to various fields, and Gang neurons show excellent performance in the corresponding fields. Interesting things: (1) The computational complexity of dendrite modules (Wi,i-1Ai-1 ○ Ai-1) connected in series is far lower than Horner’s method. Will this speed up the calculation of basic functions in computers? (2) The range of sight of animals has a gradient, but the convolution layer does not have this characteristic. This paper proposes receptive fields with a gradient. (3) The networks using Gang neurons can delete Fully-connected Layer. In other words, the parameters in Fully-connected Layers are assigned to a single neuron, which reduces the parameters of a network for the same mapping capacity. (4) ResDD(ResDD modules+One Linear module) can replace the current ANNs’ Neurons. ResDD has controllable precision for better generalization capability. Gang neuron code is available at https://github.com/liugang1234567/Gang-neuron.

“© 20xx IEEE.Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes,creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.” DOI: 10.1109/TPWRD.2020.3003250 Abstract: When a lightning strikes the top of a transmission line tower or shield wires, electromagnetic waves propagate through the tower back and forth, increasing the voltage across insulator strings. ‡is can eventually lead to a back-flƒashover (BF), which may cause damage to equipment or costly power outages. To calculate the over-voltages and predict the probability of a BF, an accurate model of the tower and its grounding system is needed in electromagnetic transient (EMT) type simulators. ‡There are a number of theoretical models for the equivalent circuit of a transmission tower. However, they either are not accurate enough or they are derived for a certain type of transmission tower, which limits their applicability. Numerical electromagnetic analyses have less simplifications compared to the theoretical solutions and are by far less expensive than field measurements. They also have the flexibility to analyze any type of tower. In this paper, the direct method for the measurement of tower impedance is implemented by NEC4 and applied to a 400-kV double circuit tower with all its details. Th‡e process of obtaining the wire network of the tower used in this paper is completely automated and it can be applied to any other type of transmission tower. Th‡e results of the numerical simulations are compared to those obtained with existing tower models. Th‡e developed model in this paper is capable of considering all the details of the tower and including the \eurofinite resistance of the ground and grounding electrodes.

This paper, discusses the implementation of RSA (Rivest–Shamir–Adleman) algorithm of encrypting and decrypting data. This is an assymetric algorithmic approach, which implies that two different keys, one public and another private have been used to accomplish the purpose. This paper also extends the application of RSA algorithm by choosing random prime numbers for every byte of transferred data. The programs involved are entirely based on the microprocessor INTEL 8085.

According to WHO, COVID-19 is an infectious disease and has a significant social and economic impact. The main challenge in ?fighting against this disease is its scale. Due to the imminent outbreak, the medical facilities are over exhausted and unable to accommodate the piling cases. A quick diagnosis system is required to address these challenges. To this end, a stochastic deep learning model is proposed. The main idea is to constrain the deep representations over a gaussian prior to reinforce the discriminability in feature space. The model can work on chest X-ray or CT-scan images. It provides a fast diagnosis of COVID-19 and can scale seamlessly. This work presents a comprehensive evaluation of previously proposed approaches for X-ray based disease diagnosis. Our approach works by learning a latent space over X-ray image distribution from the ensemble of state-of-the-art convolutional-nets, and then linearly regressing the predictions from an ensemble of classifi?ers which take the latent vector as input. We experimented with publicly available datasets having three classes { COVID-19, normal, Pneumonia. Moreover, for robust evaluation, experiments were performed on a large chest X-ray dataset with fi?ve different very similar diseases. Extensive empirical evaluation shows how the proposed approach advances the state-of-the-art.

“This paper is a postprint of a paper submitted to and accepted for publication in IET Generation, Transmission & Distibution (GTD) and is subject to Institution of Engineering and Technology Copyright. The copy of record is available at the IET Digital Library.” Abstract: Grounding electrodes have an important role in electric power transmission and distribution systems. They are used to prevent excessive hazardous voltages between metallic structures and ground in the case of system faults or lightning surges. It is important that they provide a low impedance path for the current in to the ground. The electrical properties of soil, which vary substantially with geographical location and time of year, affect the process considerably along with the properties of the grounding electrode itself, such as its dimensions. In order to have an accurate estimation of the developed overvoltages and the backflashover rate of the transmission lines due to a lightning strike, one has to take into account the effect of the value of the soil electrical parameters, such as the electrical conductivity and dielectric constant. This paper investigates the high frequency behavior of the grounding electrodes by solving a full-wave electromagnetic problem using the Finite Element Method (FEM). The focus is on taking into account the effect of the variation of soil relative permittivity which has been neglected in the previous studies of the grounding systems. This allows an evaluation of the response of grounding systems due to seasonal changes and specifically change of the water content of the soil, which would cause its electrical properties to vary significantly. This study demonstrates the importance of considering the variation of relative permittivity of the soil especially in the modeling of electrodes buried in highly resistive soil.

The 3rd Generation Partnership Project (3GPP) specification of the fifth generation (5G) New Radio (NR) allows for a highly scalable and flexible radio access technology to cater to network operators with different requirements. Such scalability and flexibilities in network configurations inevitably translate to complications in the design and implementation of 5G-NR systems. Radio access in 5G-NR is much more complex and involved than its predecessor, 4G long term evolution (LTE) and LTE-Advanced technology. Therefore, the 5G-NR specifications turn out to be quite dense. Specifically, the specifications are concise, design motivations rarely explained, and the information can be convoluted or distributed across several documents. Moreover, there are several key design details associated with the access layer procedures for any given physical layer channel, which are often omitted in the specifications. For example, design motivation aspects of initial access channels or signal generation can be quite difficult to follow or understand in 5G-NR. In this paper, all the design details associated with initial access channels and signal generation in 5G-NR specifications are laid out. The contributions of the paper are three folds. First, the design details and justifications associated with both downlink and uplink access channels are presented along with signal generation details. Secondly, receiver design aspects of NR PRACH short formats are discussed. Lastly, PRACH receiver implementation aspects and performance reports from different network operators are presented and compared with 3GPP specified Radio Performance and Protocol aspect requirements[1] for millimeter wave (mmW) access. The work in this paper is of significant value to researchers and system engineers looking to design and build initial access algorithms as part of 5G NR systems.   [1] Radio Performance and Protocol aspect requirements are specified by the 3GPP Radio Access Network working group 4, also known as RAN4.