The accurate calculation of channel electrostatics parameters, such as charge density and potential, in ultra-thin body (UTB) devices requires self-consistent solution of the Poisson’s equation and the full band structure, which is channel material and thickness dependent. For cubic crystals like silicon, the semi-empirical sp3d5s* tight-binding (TB) model is preferred in device simulations, over the density functional theory, to obtain the full band structure because of being computationally less intensive and equally accurate. However, the computational time of the TB model scales non-linearly with the channel thickness and becomes cumbersome for silicon, beyond 5 nm, primarily because of the increasing size of the TB hamiltonian that needs to be solved over the entire k-space, in the irreducible Brillouin zone. In this work, we precisely identify those k-points corresponding to the energies close to the band minima, where the Fermi-Dirac probability significantly affects electrostatics parameters. This enables us to demonstrate a computationally efficient approach based on solving the hamiltonian only on those reduced number of k-points. The rigorous benchmarking of the channel electrostatics parameters obtained from this approach is performed with results from accurate full band structure simulations showing excellent agreement over a wide range of channel thicknesses, oxide thicknesses, device temperatures and different channel orientations. By showing that the approach presented in this work is computationally efficient, besides being accurate, regardless of the number of atomic layers, we demonstrate its applicability for simulating UTB devices.

This letter introduces the concept of antisparse Blind Source Separation (BSS), proposing a suitable criterion based on the $\ell_\infty$ norm to explore the antisparsity feature. The effectiveness of the criterion is theoretically demonstrated and it is also evaluated by computational simulations, which consider up to ten distinct sources with different correlation levels. Moreover, we simulated a scenario in wireless communication with binary sources, comparing our approach to the Constant Modulus algorithm. Both the theoretical and the simulation results highlight the potentiality of using antisparsity as a prior in BSS.

A high-resolution two-dimensional (2D) electrical impedance tomography (EIT) system requires a larger number of electrodes and a finer mesh than its traditional counterpart. This increases the required number of measurements and, in turn, the amount of computation for the image reconstruction. Given the inverse and ill-posed nature of the EIT systems, they require a high signal-to-noise ratio (SNR) acquisition system as well as a high-precision hardware accelerator platform. In this paper, we present a field programmable gate array (FPGA)-based acquisition system with a tunable single-frequency current source that can reach an acquisition speed of more than 500 and 2400 frames per second (fps) for an excitation signal frequency of 500 kHz using 32 and 16 electrodes, respectively. The data processing and reconstruction are carried out using the most recent embedded Graphical Processing Unit (GPU, Nvidia Jetson Orin) by utilizing multiple Cuda cores to perform parallel high-speed 2D image reconstruction. Five different algorithms, namely linear back projection (LBP), Tikhonov regularization (TK), one-step Gauss Newton (GN), Landweber (LW), and iterative Tikhonov (ITK), were used for investigation. A gain in speed-up of at least 4 times was observed over the traditional implementations on recent general-purpose computers (PCs). Extensive experiments indicate that the proposed system can yield a throughput of more than 2500 fps for a 16-electrode system with around 8192 mesh elements. This paves the way for EIT systems to be potentially used in high-speed imaging applications as well as in 3D EIT applications which involve even larger amount of mesh elements.

We propose a BlackBox Counterfactual Explainer that is explicitly developed for medical imaging applications. Classical approaches (e.g., saliency maps) assessing feature importance do not explain how and why variations in a particular anatomical region are relevant to the outcome, which is crucial for transparent decision making in healthcare application. Our framework explains the outcome by gradually exaggerating the semantic effect of the given outcome label. Given a query input to a classifier, Generative Adversarial Networks produce a progressive set of perturbations to the query image that gradually changes the posterior probability from its original class to its negation. We design the loss function to ensure that essential and potentially relevant details, such as support devices, are preserved in the counterfactually generated images. We provide an extensive evaluation of different classification tasks on the chest X-Ray images. Our experiments show that a counterfactually generated visual explanation is consistent with the disease’s clinical relevant measurements, both quantitatively and qualitatively.

In this work, we describe step by step a complete workflow to apply machine learning (ML) classification for chipless RFID tag identification, covering: i) the tag implementation criteria for circular ring resonator (CRR) arrays for ML interoperability; ii) the data collection procedure to get a sufficiently representative dataset of real measurements; iii) the ML techniques to visualize the data and reduce its dimensionality; iv) the evaluation of the ML classifier to ensure high accuracy predictions on new measurements; and v) a thresholding scheme to increase the certainty of the predictions. The differences of the tags’ frequency responses are maximized by optimizing the Hamming distance between the tag identifiers and by controlling the radar cross section (RCS) level of each CRR array. We show on two scenarios, fixed and flexible range (up to 160 cm), that the proposed workflow can achieve perfect accuracy in most cases for the identification of four tags.

139 conductors were manufactured for the magnetic system of the ITER toroidal field coils in total by 6 different teams based on 5 different suppliers of Nb3Sn wires. Despite the fact that all conductors and superconducting wires were manufactured according to the same criteria, there were some differences in the manufacturing technologies of the components themselves, and in the result of the acceptance tests. Samples of some of the conductors passed acceptance tests, where the dependence of Tcs on the number of cycles of inputting the operating current into the samples when they were in a magnetic field was determined. The results of these tests showed that there are not only quantitative, but also qualitative differences in the Tcs (N) dependence. These differences are also observed within the same supplier. It would be useful to understand which parameters most strongly contribute to Tcs (N) in order to possibly improve this characteristic. For this purpose, the article provides a statistical analysis of the test results of 49 samples of conductors. The results are being discussed.

The paper addresses the one-class classification (OCC) problem and advocates a one-class multiple kernel learning (MKL) approach for this purpose. To this aim, based on the Fisher null-space one-class classification principle, we present a multiple kernel learning algorithm where an $\ell_p$-norm constraint ($p\geq1$) on kernel weights is considered. We cast the proposed one-class MKL task as a min-max saddle point Lagrangian optimisation problem and propose an efficient method to solve it. An extension of the proposed one-class MKL approach is also considered where several related one-class MKL tasks are learned concurrently by constraining them to share common kernel weights. An extensive assessment of the proposed method on a range of data sets from different application domains confirms its merits against the baseline and several other algorithms.

Time series analysis and forecasting of stock market prices has been a very active area of research over the last two decades. Availability of extremely fast and parallel architecture of computing and sophisticated algorithms has made it possible to extract, store, process and analyze high volume stock market time series data very efficiently. In this paper, we have used time series data of the two sectors of the Indian economy – Information Technology (IT) and Capital Goods (CG) for the period January 2009 – April 2016 and have studied the relationships of these two time series with the time series of DJIA indices, NIFTY indices and the US Dollar to Indian Rupees exchange rate. We established by graphical and statistical tests that while the IT sector of India has a strong association with DJIA indices and the Dollar to Rupee exchange rate, the Indian CG sector exhibits a strong association with the NIFTY indices. We contend that these observations corroborate our hypotheses that the Indian IT sector is strongly coupled with the world economy whereas the CG sector of India is the reflection of India’s internal economic growth. We also present several models of regression between the time series which exhibit strong association among them. The effectiveness of these models have been demonstrated by very low values of their forecasting errors.

In this paper, an intelligent reflecting surface (IRS) aided uplink (UL) rate-splitting multiple access (RSMA) system is investigated for dead-zone users where the direct link between the users and the base station (BS) is unavailable and the UL transmission is carried out only through IRS. In the considered RSMA system, a message of each user is split into several sub-messages and each part contributes to the rate of that user and depending upon split proportions BS decodes them using appropriate decoding order. The problem of sum-rate maximization is formulated to jointly design the optimal power allocation at each UL user, passive beamforming at the IRS under optimal decoding order of sub-messages. Due to non-convexity and discrete non-linear programming of the formulated problem, the original problem is intractable and hence, we decouple the problem into different sub-problems in which the problems of power allocation and passive beamforming are alternatively solved under using successive convex approximation and Riemaniann conjugate gradient algorithms, respectively. Moreover, the decoding order strategy is analytically derived which confirm that the optimal decoding order strategy depend upon decreasing order of channel gain of users and increasing order of split proportions of sub-messages. Later, the unified solution based on blockcoordinate descent (BCD) algorithm is proposed. Simulation results validate that the proposed decoding order scheme attains performance closer to the optimal solution with low computational complexity. Moreover, the proposed IRS aided RMSA system outperforms the system with non-orthogonal multiple access (NOMA) and orthogonal multiple access (OMA) schemes in terms of achievable sum-rate throughput.

The drivetrain flexibility of industrial robots limits their accuracy. To open up new areas of application for industrial robots, an increased dynamic path accuracy has to be obtained. Therefore, this paper addresses this issue by a gain-scheduled drive-based damping control for industrial robots with secondary encoders. For this purpose, a linear parameter-varying (LPV) model is derived as well as a system identification method is presented. Based on this, a gain-scheduled drive-based LPV damping control design is proposed, which guarantees stability and performance under variation of the manipulator configuration. The control performance of the approach is experimentally validated for the three base joints of a KUKA KR210-2 industrial robot. The approach realizes a trade-off between ease of implementation and control performance as well as robustness.

The Mobile Edge computing (now known as multi-access edge computing) has been widely recognized as a key enabler for new latency sensitive applications on resource constrained mobile devices. The objective to offload a computationally intensive task to a cloud server has been extensively researched over the years. These research endeavours have been in general aimed at reducing system’s energy consumption and/or latency. In this paper, we attempt to examine how profitable computation offloading is from a service provider’s perspective. The joint optimization of radio and computational resources along with offloading decisions result in a mixed integer non-linear optimization problem which belong to the class of NP hard problems. To counter this challenge, we decouple the offloading decision from the resource allocation problem. At first, an approximate optimal offloading decisions are found using evolutionary algorithms like genetic algorithm and binary particle swarm optimization algorithm where the objective function is approximately calculated using machine learning based model rather than actually solving the optimization problem. Then the combination of selected mobile terminals for offloading further goes as input to resource allocation optimization problem. According to the simulations performed, the proposed evolutionary algorithms outperform spectrum efficiency based offloading algorithm in terms of profitability and have relatively lower execution times. Also, the impact of resource availability on profitability of offloading has been examined.

Software-defined Networking (SDN) can facilitate the deployment of deterministic algorithms with stringent Quality of Service (QoS) requirements that pave the way for commodity Real-time (RT) networks. However, current QoS approaches are conservative and under-provision the network, that becomes a bottleneck when scaling or upgrading infrastructure. In this paper, we argue that the use of fixed priorities for flows – a common practice in RT networks, is the root-cause of this issue. We develop algorithms that use the global view provided by SDN to develop on-demand, variable priority schemes. Evaluation shows that this “correct by construction” approach increases network utilization while still meeting the necessary QoS requirements, thereby improving network robustness under high utilization.

Data-driven approaches for molecular diagnostics are emerging as an alternative to perform an accurate and inexpensive multi-pathogen detection. A novel technique called Amplification Curve Analysis (ACA) has been recently developed by coupling machine learning and real-time Polymerase Chain Reaction (qPCR) to enable the simultaneous detection of multiple targets in a single reaction well. However, target classification purely relying on the amplification curve shapes currently faces several challenges, such as distribution discrepancies between different data sources of synthetic DNA and clinical samples (i.e., training vs testing). Optimisation of computational models is required to achieve higher performance of ACA classification in multiplex qPCR through the reduction of those discrepancies. Here, we proposed a novel transformer-based conditional domain adversarial network (T-CDAN) to eliminate data distribution differences between the source domain (synthetic DNA data) and the target domain (clinical isolate data). The labelled training data from the source domain and unlabelled testing data from the target domain are fed into the T-CDAN, which learns both domains’ information simultaneously. After mapping the inputs into a domain-irrelevant space, T-CDAN removes the feature distribution differences and provides a clearer decision boundary for the classifier, resulting in a more accurate pathogen identifi- cation. Evaluation of 198 clinical isolates containing three types of carbapenem-resistant genes (blaNDM , blaIMP and blaOXA-48 ) illustrates a curve-level accuracy of 93.1% and a sample- level accuracy of 97.0% using T-CDAN, showing an accuracy improvement of 20.9% and 4.9% respectively, compared with previous methods. This research emphasises the importance of deep domain adaptation to enable high-level multiplexing in a single qPCR reaction, providing a solid approach to extend qPCR instruments’ capabilities without hardware modification in real- world clinical applications.

# Machine learning Classifiers for prediction of Pathway module & it classes We use SMILES representation of query molecules to generate relevant fingerprints, which are then fed to the machine learning classifiers ETC for producing binary labels corresponding pathway module & its classes. The details of the works are described in our paper. A dataset of 6597 downloaded from KEGG, 4612 compounds either belong or not to Pathway module in metabolic pathway the remaining 1985 compounds belong to module classes prediction problems ### Requirements *Chemoinformatics tools * Python * scikit-learn * RDKit * Jupyter Notebook ### Usage We provide two folder containing Classifiers files,grid search for optimization of hyperparameters, and datasets(module, module classes

This paper presents a virtual thermal sensor for real-time monitoring of electronic packages in a totally enclosed system.

With the acquisition of massive condition monitoring data, how to realize real-time and efficient intelligent fault diagnosis is the focus of current research. Inspired by the ideas of compressed sensing (CS) and deep extreme learning machines (DELM), a data-driven lightweight framework is proposed to accelerate intelligent fault diagnosis. The integrated framework contains two modules: data sampling and fault diagnosis. Data sampling module projects the intensive original monitoring data into lightweight compressed sampling data non-linearly, which can effectively reduce the pressure of transmission, storage and calculation. Fault diagnosis module digs deeply into the inner connection between the compressed sampled signal and the fault types to realize accurate fault diagnosis. This work has three meaningful points. First, we believe that the bearing vibration signal is not strictly sparse in the transform domain. Second, we verified that the sparse signal after compressed sampling can be directly used for fault diagnosis without being reconstructed. Third, adding a kernel function to the DELM can perfectly map the low-dimensional inseparable features after compressed sampling to the high-dimensional space non-linearly to make it linearly separable and thus improve the classification accuracy

This paper introduces a novel transient simulation scheme and its extension for stability analysis of distribution systems with converter-based distributed generation. Efficiency and accuracy of the simulation scheme with extension are investigated. It is demonstrated that millisecond-level step sizes are applicable to efficient electromagnetic transient simulation for stability analysis of unbalanced power systems while maintaining satisfactory accuracy. The paper also supports the extension of the simulation scheme by showing that the extension has only an insignificant impact on the accuracy but significantly enhances the efficiency. This paper has been accepted by the 2021 IEEE IAS Annual Meeting.

Line waves are modes that propagate along a line connecting two-dimensional impedance surfaces with dual electromagnetic responses. In general, line waves can be implemented using complementary impedances or impedances with a gain-loss effect. On the other hand, if the two impedance surfaces are non-complementary, they can support a different type of mode called a “quasi-line wave”. We propose a different configuration to implement quasi-line waves. Indeed, non-dual structures with purely inductive impedances are used in order to drive the mode. We investigate the implementation of quasi-line modes using multilayer graphenes. This mode has a wide bandwidth in the terahertz range and a significant propagation length. The proposed structure guides one-way modes in which the fields are highly concentrated around the edges of two inductive impedance structures. Furthermore, incorporating graphene into the proposed waveguide allows for control of bandwidth and transmission characteristic of the waveguide.