Search

In this paper, we study the state-dependent Gaussian wiretap channel with action encoder and noncausal state at the transmitter, which is also called the action-dependent dirty paper wiretap channel. Two lower bounds on the secrecy capacity of this model are provided, and we find that the two bounds have their advantages and disadvantages, i.e., none of the two always performs better than the other. In addition, we find that taking action on the state helps to increase the achievable secrecy rate of the dirty paper wiretap channel.

Economically paper and pulp mills are important but at the same time it is facing problems due to pollutants that are produced during the process. In this paper we will discuss methods of waste water treatment in the industry. This industry use excessive amount of water because it is used in nearly each and every step of the process as a result large quantity of waste water and residual sludge waste is produced leading to the problems of wastewater treatment, discharge and sludge disposal but advanced technologies with new equipment with the help of some innovative strategies are making it possible for treating wastewater for its reuse by presenting solutions to the industries and helping them in choosing suitable and affordable methods for management of waste

Cyber networks are extremely non-deterministic, complex systems. To address this, we must develop foundational research protocols to enable reproducible cyber experiments that can systematically uncover deep understanding of a cyber system's security posture. One core tenant of this approach is to have a test environment to enable a space to create and test hypotheses about systems and reason about results. To date, this has generally been done through cyber testbeds or cyber ranges. National infrastructure supported by various government agencies have all created multi-million dollar ranges, and support other national infrastructure such as the National Cyber Range (NCR) and the Regional Service Delivery Points (RSDP). The thrust of this paper was based on multi-year studies, the culmination of which uncovered gaps and challenges associated with using various national infrastructures to represent a multitude of complex heterogeneous systems. Ranges, such as the NCR, have experiment life-cycle processes to take a cyber experiment from inception to analysis. However, our position is that processes used are not sufficient to address gaps and challenges. In this position paper, we review current range experiment methodologies and our observations of other considerations that should require inclusion.

Due to the long-term operation, the transformer oil-paper insulation system will gradually experience aging and degrading, resulting in the insulation failure. And the insufficient research about the accurate diagnosing method of aging condition will severely threaten the security and stability of the transformer. At present, the diagnostic criteria for the aging of transformer insulation is only limited to the degree of polymerization of the insulating paper/pressboard and furfural in the oil without a comprehensive diagnostic scheme. Based on the existing theory, for the oil-paper insulation system after undergoing the accelerated thermal aging test, the degree of polymerization of the insulation paper, frequency domain dielectric properties, furfural, CO, CO 2 were firstly tested in the laboratory. Then, principal component analysis (PCA) was used to integrate the measured multiple parameters and obtain a comprehensive evaluating index of the insulation aging condition, and the establishment of the degree of polymerization and the evaluating index was conducted, which was in good linear relationship. According to the evaluating index and the quantitative relationship with the degree of polymerization, a diagnostic method that can be applied to the aging assessment of transformer insulation in the field is put forward, which makes up for the deficiency of the current diagnosis method only including single assessment factor and is of certain practical significance.

Even though AI/ML is of strategic importance for many companies, it is not easy to come up with an AI life cycle in a multi-cloud world that is being increasingly embraced. In this paper, we present our position on an AI strategy that is future proof and does not force vendor lock in.

This effort presents an analysis of a system intended for wireless crack sensing and based on zero-power harmonic transponders. The sensors, which are fabricated by using a paper substrate, are suitable for both structural health and supply-chain monitoring. In the text, different techniques to recover information from the proposed tags are firstly explored, with a particular focus on the potentialities and trade-offs of each approach. Then, the impact of the presence of multiple-nodes on the performance of the system is considered and, finally, some proposals for crack localization in presence of multiple tags are explored and experimentally evaluated.

Most existing studies and public datasets for handwritten Chinese text recognition are based on the regular documents with clean and blank background, lacking research reports for handwritten text recognition on challenging areas such as educational documents and financial bills. In this paper, we focus on examination paper text recognition and construct a challenging dataset named examination paper text (SCUT-EPT) dataset, which contains 50 000 text line images (40 000 for training and 10 000 for testing) selected from the examination papers of 2 986 volunteers. The proposed SCUT-EPT dataset presents numerous novel challenges, including character erasure, text line supplement, character/phrase switching, noised background, nonuniform word size, and unbalanced text length. In our experiments, the current advanced text recognition methods, such as convolutional recurrent neural network (CRNN) exhibits poor performance on the proposed SCUT-EPT dataset, proving the challenge and significance of the dataset. Nevertheless, through visualizing and error analysis, we observe that humans can avoid vast majority of the error predictions, which reveal the limitations and drawbacks of the current methods for handwritten Chinese text recognition (HCTR). Finally, three popular sequence transcription methods, connectionist temporal classification (CTC), attention mechanism, and cascaded attention-CTC are investigated for HCTR problem. It is interesting to observe that although the attention mechanism has been proved to be very effective in English scene text recognition, its performance is far inferior to the CTC method in the case of HCTR with large-scale character set.

As Machine Learning (ML) becomes pervasive in the era of artificial intelligence, ML specific tools and frameworks are required for architectural research. This paper introduces NeuroMeter, an integrated power, area, and timing modeling framework for ML accelerators. NeuroMeter models the detailed architecture of ML accelerators and generates a fast and accurate estimation on power, area, and chip timing. Meanwhile, it also enables the runtime analysis of system-level performance and efficiency when the runtime activity factors are provided. NeuroMeter’s micro-architecture model includes fundamental components of ML accelerators, including systolic array based tensor units (TU), reduction trees (RT), and 1D vector units (VU). NeuroMeter has accurate modeling results, with the average power and area estimation errors below 10% and 17% respectively when validated against TPU-v1, TPU-v2, and Eyeriss.Leveraging the NeuroMeter’s new capabilities on architecting manycore ML accelerators, this paper presents the first in-depth study on the design space and tradeoffs of “Brawny and Wimpy” inference accelerators in datacenter scenarios with the insights that are otherwise difficult to discover without NeuroMeter. Our study shows that brawny designs with 64x64 systolic arrays are the most performant and efficient for inference tasks in the 28nm datacenter architectural space with a 500mm2 die area budget. Our study also reveals important tradeoffs between performance and efficiency. For datacenter accelerators with low batch inference, a small $(\sim 16$%) sacrifice of system performance (in achieved Tera OPerations per Second, aka TOPS) can lead to more than a 2x efficiency improvement (in achieved TOPS/TCO). To showcase NeuroMeter’s capability to model a wide range of diverse ML accelerator architectures, we also conduct a followon mini-case study on implications of sparsity on different ML accelerators, demonstrating wimpier accelerator architectures benefit more readily from sparsity processing despite their lower achievable raw energy efficiency.

In this invited paper, applications for Machine Learning (ML) in several areas of semiconductor manufacturing and test are reviewed and potential opportunities are discussed.

Microfluidic engineering technology has been widely used for implementing lab-on-chip (LOC) type point of care (POC) devices since its origins in the 1990s. LOC devices offer some distinct advantages such as the possibility of reducing the quantity of valuable samples or reagents needed and the shortening of the detection times, both of which result from their compact structure [1]. Paper-based microfluidic devices, which are regarded as a low-cost alternative to conventional POC diagnostics tools, have also been popularly studied in the last few decades, and a huge number of advantages have been explored, such as low-cost, mass production, disposability, being equipment free etc. [2] However, some disadvantages or limitations are also apparent, such as issues with the control of flow rate, multiplexed detection on the same device footprint and further reduction of size [2]. As a result, 3D microfluidic paper analytical devices, as an alternative solution, have been proposed in recent years, which enable distribution of fluids in both lateral and vertical directions.

Nowadays, there is a rapidly increasing number of conferences and journals in computer science that make a lot of challenges for researchers to find an appropriate venue to submit their scientific work. There is a need for a recommendation system that can support researchers for a better process of paper submission. In this paper, we present an efficient approach for building such a recommendation model by using embedding methods, Global Vector (GloVe) 1 created by Pennington et al. [1] and FastText 2 proposed by Facebook [2], Convolutional Neural Network (CNN) [3], and LSTM. The experimental results show that the combination of CNNs and FastText, CNN + FastText, can achieve the best performance in terms of the Top 1 Accuracy compared with other techniques, including the S2RSCS model, as presented in [4]. Moreover, the performance by using GloVe or FastText is much better, faster, and more stable than S2RSCS in most cases.

In the context of papers used in the graphic arts, including silver gelatin, inkjet, and wove papers, prior work has studied measures of texture similarity for purposes of classifying such papers. The majority of prior work has been based on classical image processing approaches such as Fourier, wavelet, and fractal analysis. In this work, recent advances in deep learning are used to develop a texture similarity approach for measuring paper texture similarity. Since the available datasets generally lack labels, the convolutional neural network is trained using triplet loss to minimize the feature distance of tiles from the same image while simultaneously maximizing the feature distance of tiles drawn from different images. The approach is tested on three paper texture image databases considered in prior works and the results suggest the proposed approach achieves state-of-the-art performance.

Anti-counterfeiting technique is a research hotspot in information security domain. Common techniques are usually based on artificial physical cloneable feature which is forgeable. In this paper, we proposed an anti-counterfeit method based on microscopic translucent texture of paper (MTTP). MTTP is image of internal fiber pattern of paper (IFPP) captured by microscope under translucent light source. The IFPP has property of uniqueness and randomness, and the fiber scale of IFPP is at sub-micron or micron level. Then the IFPP is unable to be forged. Furthermore, fiber texture in IFPP is stable under disturbances (including illumination intensity change and surface damage), which guarantees the stability of the MTTP. Experiment results show that the authentication accuracy achieve 100% in our dataset (totally 13,4550 pairs of MTTPs) and the false reject rate achieves the order of magnitude of 10-136 in ideal condition and 10-59 even in condition of illumination change or surface damage. Above results indicate that our method has rather high security and reliability.

Prior work on texture analysis of historic, photographic papers has focused primarily on measures of texture similarity. However, automated grouping or clustering of photographic paper textures in a way that is meaningful to art conservators remains an open problem. In this work a deep learning approach to automated classification is presented, for clusters derived from a human sorting experiment conducted by 19 art conservators and paper experts and subsequently extended through crowd-sourcing. The proposed approach uses a deep convolutional neural network, and results are presented on the performance in automatically classifying images when compared to human experts.

We develop and commercialize autonomous machines, such as logistic robots and self-driving cars, around the globe. A critical challenge to our—and any—autonomous machine is accurate and efficient localization under resource constraints, which has fueled specialized localization accelerators recently. Prior acceleration efforts are point solutions in that they each specialize for a specific localization algorithm. In real-world commercial deployments, however, autonomous machines routinely operate under different environments and no single localization algorithm fits all the environments. Simply stacking together point solutions not only leads to cost and power budget overrun, but also results in an overly complicated software stack.This paper demonstrates our new software-hardware co-designed framework for autonomous machine localization, which adapts to different operating scenarios by fusing fundamental algorithmic primitives. Through characterizing the software framework, we identify ideal acceleration candidates that contribute significantly to the end-to-end latency and/or latency variation. We show how to co-design a hardware accelerator to systematically exploit the parallelisms, locality, and common building blocks inherent in the localization framework. We build, deploy, and evaluate an FPGA prototype on our next-generation self-driving cars. To demonstrate the flexibility of our framework, we also instantiate another FPGA prototype targeting drones, which represent mobile autonomous machines. We achieve about $2 \times$ speedup and $4 \times$ energy reduction compared to widely-deployed, optimized implementations on general-purpose platforms.

Paper-based digital microfluidic biochips (PDMFBs) are becoming highly effective among the microfluidic platforms due to its low-cost and in-place fabrication. The designed electrodes and wiring can be fabricated on a piece of paper by an inkjet printer and conductive ink containing carbon nanotube particles (CNTs). However, due to induced control interference, the wires cannot pass by an arbitrary electrode. Each wire that is to be routed possesses its conflict electrode group, which must be avoided for a feasible droplet movement. This paper presents a fluid-control co-design considering several important cost-driving issues like minimization of schedule length, control pin count, and wirelength, together with congestion-free and conflict-free wiring. Observably, design gaps exist among the sub-tasks of the fluid-level, control-level, and fluid-control as a whole, due to their separate considerations. It indeed introduces many design cycles lengthening the design process, and thus increases the overall cost. In this context, this work integrates the sub-tasks as a prior need to obtain a low cost and efficient platform. Several benchmarks have been studied to evaluate the performance.

Physical synthesis has emerged as a core component in a modern circuit design flow. Large-scale optimization problem is often involved in the process, which requires substantial efforts to solve and no optimality is guaranteed. Reinforcement learning provides one direction to deal with the above issue by automatically acquiring knowledge through experience, which has shown great success in various applications. In this paper, we introduce the foundation of and the progress in reinforcement learning, and review some recent approaches in applying reinforcement learning to physical synthesis. We hope to inspire more work and to see more talented ideas in this field.

This paper considers a cell-free massive MIMO system with multiple-antenna access points and single-antenna users. The APs use conjugate beamforming to beamform the data to all users in the network. Total energy efficiency maximization is investigated. This optimization problem is nonconvex and thus difficult to solve. Existing solutions are based on second-order optimization methods in connection with convex approximations. These methods have been shown to perform very well but their complexity does not scale favorably with the network size. To tackle this issue, in this paper we propose to use a first-order method for nonconvex programming to our energy efficiency problem. Compared to the second-order methods, the proposed method achieves the same performance, while its run time is much faster. Thus, it is considered as a feasible solution for resource allocation in cell-free massive MIMO systems.

We propose a passive touch sensor that can be implemented on semi-conductive materials, such as paper. This sensor detects electric currents leaked from a human body when touched with bare fingers. Such electric currents are mostly produced by AC power lines in the environment. We tested the potential of the sensor for gesture recognition by using a paper sheet painted with a hydrogel ink. A gesture classification algorithm combining the $k$ -means method and Mahalanobis' generalized distance could successfully classify 90% of the gestures drawn on the paper with a fingertip. Gestures including multi-directional lines and circles were nearly perfectly classified whereas digits, such as 2 and 8 exhibited difficulties with classification. The hydrogel ink can turn other non-conductive materials comprising woods and plastic into touch sensors, and such hum-based touch sensors are expected to find practical applications in gesture recognition.

Facial Expressions play a vital role in the process of recognizing emotions and also to have non-verbal communication as well as in identifying people. Emotional Recognition turns out to be very important in everyday life just next to the tone of voice. Novel applications in Human-Computer Interaction (HMI) have been enabled by this and in many other areas. Inevitably most of the recent research on this area focuses on Convolutional Neural Networks (CNN) for extraction of features and inference from those features. In this paper we used CNN algorithm along with Histogram of gradients (HOG) features for higher accuracy. We perform emotional analysis of lecturers from Impartus videos. By detecting their emotions from a sequence of videos during their coursework, we evaluate the feedback they are expected to receive at the end of the duration of their course.

Diabetic Retinopathy (DR) is a polygenic disorder issue that affects human eyes. Bruise to the blood vessels of the photosensitive tissue of the retina causes this complication. It’s most frequent in patients who had diabetes for more than ten years. This downside is going on in several individuals worldwide. However, the number of medical practitioners and also the tools needed for the detection of DR are very less for serving the mass population. In this paper, we have proposed DRDNet (Diabetic Retinopathy Diagnosis Network), a neural network framework based on capsule networks (CapsNets) for DR diagnosis. Experiments on a dataset with 1,265 images demonstrate that CapsNet shows better accuracy and convergence behavior for the complex data than the state-ofthe-art techniques. The proposed DRDNet performs with an overall accuracy of 80.59% for five class, as compared to the closest competitor with an accuracy of 75.83%. We performed a study on a mixed dataset for two class and found that testing accuracy was 80.59%. We have also done training on a two class model and testing on other unseen datasets. Moreover, we observed that DRDNet has much higher confidence for the predicted probabilities as compared to other state-of-the-art techniques.

By integrating both energy harvesting and backscatter communication technologies, so-called ‘battery-free tag’ emerges as a promising solution to the energy related issues in IoT. However, such tags present new challenges due to the unstable energy supply and excitation signals, which are critical to realize successful backscatter communications. In this paper, we present a novel system, denoted as ecUWB, which enables robust communication for battery-free tags where unstable WiFi signals act as the unified source for both energy supply and excitation signals.At tag side, we propose a charging-transmission division scheme to achieve better signal utilization, and introduce a re-transmission mechanism for possible excitation signal interruption. At receiver side, we implement a simply method which is based on DBSCAN (Density-Based Spatial Clustering of Applications with Noise) to predict the uncontrollable signals, and propose an energy-centric tag scheduling method according to the tag energy estimation results. Extensive experiments are carried out on customized tags and NI USRP platform. The results show that ecUWB outperforms the existing ones in terms of performance and efficiency under the unstable WiFi signals.

Value Prediction was proposed to speculatively break true data dependencies, thereby allowing Out of Order (OOO) processors to achieve higher instruction level parallelism (ILP) and gain performance. State-of-the-art value predictors try to maximize the number of instructions that can be value predicted, with the belief that a higher coverage will unlock more ILP and increase performance. Unfortunately, this comes at increased complexity with implementations that require multiple different types of value predictors working in tandem, incurring substantial area and power cost. In this paper we motivate towards lower coverage, but focused, value prediction. Instead of aggressively increasing the coverage of value prediction, at the cost of higher area and power, we motivate refocusing value prediction as a mechanism to achieve an early execution of instructions that frequently create performance bottlenecks in the OOO processor. Since we do not aim for high coverage, our implementation is light-weight, needing just 1.2 KB of storage. Simulation results on 60 diverse workloads show that we deliver 3.3% performance gain over a baseline similar to the Intel Skylake processor. This performance gain increases substantially to 8.6% when we simulate a futuristic up-scaled version of Skylake. In contrast, for the same storage, state-of-the-art value predictors deliver a much lower speedup of 1.7% and 4.7% respectively. Notably, our proposal is similar to these predictors in performance, even when they are given nearly eight times the storage and have 60% more prediction coverage than our solution.

In this paper, we introduce the optimal design methods of 3D NAND nvCIM [1], and then address the read disturb reliability issue. In recent years CIM [2] is widely considered as a promising solution to accelerate the DNN inference hardware. Theoretically, nvCIM can drastically reduce the power consumption by data movement because of no need to move the weights during computation. 3D NAND has the advantage of extremely low Icell (~nA), while the large ON/OFF ratio provides the capability to sum >10’000 cells together to improve the performance bandwidth and energy efficiency. We think that 3D NAND nvCIM has the potential to serve as the inference accelerator for the high-density fully- connected (FC) network which often requires high-bandwidth inputs. The read disturb property is studied. It is suggested that the "on-the-fly" calibration technique can well maintain the inference accuracy for 10-year usage.

The low carrier mobility of (printed) semiconductors and high variations of the printed/flexible electronic elements are some of the most difficult challenges for the practical realization of complex printed/flexible electronic circuits. Although contemporary circuit designs accommodate some degree of process variations in conventional silicon processes largely by negative feedback, negative feedback is largely inapplicable in printed/flexible electronics. This is because the gain of the transistors is too low in electronic circuits to realize high open-loop gain, but also because negative feedback is inapplicable to digital circuits in the traditional sense. In this paper, we describe the co-design between the first three supply chains of Printed/Flexible Electronics – the co-design between our modified semiconductor, Fully-Additive Low-Temperature All-Air Low-Variation Printed/Flexible Electronics printing process and a process-variation-tolerant digital circuit design methodology. The first two co-design supply chains pertain to a low-cost screen-printing process while the last co-design chain pertains to the Quasi-Delay-Insensitive asynchronous-logic digital design methodology (vis-a-vis synchronous-logic) that is self-timed, hence virtually tolerant to any variations. We will also delineate our measurements on digital circuit based on codesign of the aforesaid supply chains, depicting the merits of our modified semiconductor, low-variation printing process, and the ensuing functional printed digital circuit.

With fast switching GaN any parasitic gate loop inductance degrades the switching performance and may lead to false turn-on as well as gate voltage overshoot. Two approaches to overcome these challenges in driving GaN transistors are discussed in this paper. In a discrete silicon based driver, the gate loop inductance is actively utilized for a resonant gate drive approach. In a second implementation, the gate loop inductance is reduced close to zero by GaN-on-Si monolithic integration of the power transistor and the driver on one die. It includes an integrated supply voltage regulator circuit that generates the gate drive voltage out of the high-voltage switching node. The results show fast and robust switching behavior with minimal ringing.

This paper introduces the time stamped bijective MAC algorithm (bMAC), which aims at proving firmware integrity in embedded systems. bMAC computes a MAC over a set of memories, in a pseudo random way relying on a permutation. It is based on exponential permutations that induce computing time distributions. Because bMAC performs calculations on SRAM content, it can be bound to physical chips, by means Physical Unclonable Functions (PUF), extracted from SRAM.

This talk will review our recent work on a high-speed avalanche photodiode (APD). Recent results show 100-Gbit/s PAM4 40-km transmission with receiver sensitivity exceeding -14 dBm when the APD is combined with in-house SOA-integrated EML.

Photonic integrated circuit (PIC), as a promising alternative to traditional CMOS circuit, has demonstrated the potential to accomplish on-chip optical signal transmission and computations in ultra-high speed and/or low power consumption. One of the critical challenges of PIC, however, is that its scalability and robustness are limited by cascaded optical power loss and noise error. In this paper, we analyze the scalability and noise robustness challenges facing photonic integrated circuits, for two representative PIC applications: logic computing and neural networks. Automated design algorithms and learning methodologies are proposed to resolve these issues.

Tensor provides a brief and natural representation for large-scale multidimensional data by way of appropriate low-rank approximations, thus we can discover significant latent structures of complex data and generalize data representation. To date, tensor has gained tremendous success in various science and technology fields, especially in machine learning and big data applications. However, tensor computation, especially tensor decomposition, is usually expensive due to the inherent large-size characteristic of tensors, and hence would potentially hinder their future wide deployment. In this paper, we develop a hardware architecture to accelerate tensor singular value decomposition (t-SVD), which is a new tensor decomposition technique that has been successfully applied to high-dimensional data classification and video recovery. Specifically, design consideration of each key computing unit is analyzed and discussed. Then, the proposed t-SVD hardware architecture is implemented and synthesized using CMOS 28nm technology. Comparison with real-world CPU-based implementations shows that the proposed hardware accelerator is expected to provide average 14× speedup on various t-SVD workloads.

Electrowetting-on-dielectric (EWOD) is a mechanism that allows physical handling of liquids with only electrical signals, such as digitizing a liquid into tiny droplets and manipulating them on a chip, thus enabling “digital microfluidics”. As an elegantly simple platform free of pumps or valves, EWOD digital microfluidics has been attracting high research interest in the past two decades and has recently been transitioned to a few commercial products in displays and biochemistry. However, the number of labs utilizing this technology is still small, due to the difficulty in translating design intent to manufactured devices. Accordingly, in this paper, we propose a cloud-based open-source EWOD cybermanufacturing ecosystem that enables an automatic translation from user requirements to manufactured digital microfluidics. The incubation cyber ecosystem is to be accessible to a wide range of end users, allowing researchers, entrepreneurs, students, and hobbyists alike to focus on their own ideas and applications without having to master the subtleties of EWOD engineering and manufacturing. This can be thought of as an “operating system” for EWOD community, similar to Windows and macOS for people with no computer hardware background. The proposed cyber ecosystem aims to ease the process of designing and eliminate the burden of fabricating EWOD chips so that the user pool is increased and more applications found for EWOD digital microfluidics.

Video detection and classification constantly involve high dimensional data that requires a deep neural network (DNN) with huge number of parameters. It is thereby quite challenging to develop a DNN video comprehension at terminal devices. In this paper, we introduce a deeply tensor compressed video comprehension neural network called DEEPEYE for inference at terminal devices. Instead of building a Long Short-Term Memory (LSTM) network directly from raw video data, we build a LSTM-based spatio-temporal model from tensorized time-series features for object detection and action recognition. Moreover, a deep compression is achieved by tensor decomposition and trained quantization of the time-series feature-based spatio-temporal model. We have implemented DEEPEYE on an ARM-core based IOT board with only 2.4W power consumption. Using the video datasets MOMENTS and UCF11 as benchmarks, DEEPEYE achieves a 228.1× model compression with only 0.47% mAP deduction; as well as 15k× parameter reduction yet 16.27% accuracy improvement.

Neural networks are widely used in many applications ranging from classification to control. While these networks are composed of simple arithmetic operations, they are challenging to formally verify for properties such as reachability due to the presence of nonlinear activation functions. In this paper, we make the observation that Lipschitz continuity of a neural network not only can play a major role in the construction of reachable sets for neural-network controlled systems but also can be systematically controlled during training of the neural network. We build on this observation to develop a novel verification-aware knowledge distillation framework that transfers the knowledge of a trained network to a new and easier-to-verify network. Experimental results show that our method can substantially improve reachability analysis of neural-network controlled systems for several state-of-the-art tools.

Micro aerial vehicle (MAV) gradually gains plentiful attention in recent years. However, it is still a full challenge to design MAVs with high aerodynamic performance because aerodynamic efficiency of wings deteriorates sharply at low Reynolds numbers. Under this circumstance, we focus on one unconventional configuration with annular wing, which can be folded from a square paper. The annular paperplane configuration is described with geometric parameters, and the longitudinal and lateral aerodynamic characteristics are studied with RANS method.

Robustness is a key requirement for deploying a machine learning (ML) based solution. When a solution involves a ML model whose robustness is not guaranteed, ensuring robustness of the solution might rely on continuous checking of the ML model for its validity after the solution is deployed in production. Using wafer image classification as an example, this paper introduces tensor-based methods that help improve robustness of a neural-network-based classification approach and facilitate its deployment. Experiment results based on data from a commercial product line are presented to explain the key ideas behind the tensor-based methods.

This paper investigates duobinary signals and their applications in non-orthogonal multi-carrier systems to enhance spectral efficiency. In duobinary transmission schemes, the signal spectrum is reshaped by introducing controlled correlation, which can be eliminated at the duobinary decoder. For the first time, we propose the idea of combining duobinary transmission technique and the fast orthogonal frequency division multiplexing (Fast-OFDM) system with three subcarriers, with experimental results presented. The proposed system is capable of achieving three times the data rate of single-carrier ASK scheme with the same bandwidth. Results show that bit error rate (BER) performance of the proposed duobinary-based Fast-OFDM system is slightly worse than the ASK system. In addition, we also tested various 3-PSK constellation patterns designed for the duobinary signal to achieve performance improvement.

Fiber nonlinearities from Kerr effect are considered as major constraints for enhancing the transmission capacity in current optical transmission systems. Digital nonlinearity compensation techniques such as digital backpropagation can perform well but require high computing resources. Machine learning can provide a low complexity capability especially for high-dimensional classification problems. Recently several supervised and unsupervised machine learning techniques have been investigated in the field of fiber nonlinearity mitigation. This paper offers a brief review of the principles, performance and complexity of these machine learning approaches in the application of nonlinearity mitigation.

This paper outlines the importance of gaining a good knowledge of the underwater channel. The channel is discussed in terms of absorption and scattering with an explanation of the processes involved in both. Then a review of the various techniques used to model these processes is presented. Some results obtained via a Monte-Carlo simulation of the channel are presented and discussed, showing how channels with similar parameters but different scattering conditions can have very different responses.

Steganography is the art of hiding a secret message inside a publicly visible carrier message. Ideally, it is done without modifying the carrier, and with minimal loss of information in the secret message. Recently, various deep learning based approaches to steganography have been applied to different message types. We propose a deep learning based technique to hide a source RGB image message inside finite length speech segments without perceptual loss. To achieve this, we train three neural networks; an encoding network to hide the message in the carrier, a decoding network to reconstruct the message from the carrier and an additional image enhancer network to further improve the reconstructed message. We also discuss future improvements to the algorithm proposed.

In recent times, Generative Adversarial Networks (GANs) have created a lot of buzz in the research community. GANs are formulated on the zero-sum game theory, where two neural nets compete against each other. The resultant deep model is capable of generating data similar to any data distribution provided. It utilizes the adversarial learning approach and is far more capable in learning features than the traditional machine learning models. This review focusses on the origin and evolution of GANs. Firstly, the traditional GAN is explored in terms of its structure and loss functions. Then come the common challenges of training GANs. Thirdly, the review dives into numerous GAN variants and explains their improvements. The review then lists the wide variety of applications and ends with the conclusion.

Pulse shapers are widely used in particle detector applications to reduce the rise and fall times of the pulse signal and to remove noise from the frontend amplifiers. We demonstrate a low-noise, low-power pulse shaper employing an active lowpass Salley-Key biquad filter, passive highpass filter, and input/output buffers. The design process is described to generate the desired transfer characteristic while incorporating tuning knobs to adapt the shaper characteristic to the needs of different applications. Utilizing only a single 1.8-V supply, the shaper consumes 1.5 mW and achieves an input-referred noise of 17 μV rms , both the lowest among state-of-the-art. The power efficiency normalized to bandwidth and noise is nearly 1.7× better than the next best design.

Survivability is an important component in the requirements in elastic optical networks (EONs). We examine here the significance of network survivability design against multiple-link failures under dedicated protection, shared risk link group (SRLG) and bandwidth squeezing schemes. The proposed mixed integer linear programming (MILP) formulation seeks to derive some different types of protection considering routing, spectrum assignment, grooming, modulation format as well as shared risk link group constraints. The proposed MILP provides efficient survivability results and resource savings (in terms of spectrum) for a full design of modern EONs.

Quantum Approximate Optimization Algorithm (QAOA) provides approximate solution to combinatorial optimization problems. It encodes the cost function using a $p$ -level quantum circuit where each level consists a problem Hamiltonian followed by a mixing Hamiltonian. Despite the promises, few real-world applications (besides the pedagogical MaxCut problem) have exploited QAOA. The success of QAOA relies on the classical optimizer, variational parameter setting, and quantum circuit design and compilation. In this study, we implement QAOA and analyze its performance for a broader Quadratic Unconstrained Binary Optimization (QUBO) formulation to solve real-word applications such as, partially occluded object detection problem. Furthermore, we analyze the effects of above influential factors on QAOA performance. We propose a 3-level improvement of hybrid quantum-classical optimization for object detection. We achieve more than 13X execution speedup by choosing L-BFGS-B as classical optimizer at the first level and 5.50X additional speedup by exploiting parameter symmetry and more than 1.23X acceleration using parameter regression at the second level. We empirically show that the circuit will achieve better fidelity by optimally rescheduling gate operations (especially for deeper circuits) at the third level.

As the complexity of hardware (sensors, components, antennas) and software (algorithms) increases, it is practical and efficient to manage and process test configuration and data analysis as close to the testbed as possible (inline) instead of offline compute platforms. We present mmMoReEdge, a mmWave modular and reconfigurable testbed inspired by a smart edge networking and communication framework, typically found in IoT devices. In mmMoReEdge, complex signal processing is performed on the edge (local servers in close proximity) of a group of testbed nodes. mmMoReEdge offers modularity via configuration of phased-array antennas, RF front ends, ADC, and DAC, while the edge processing provides reconfigurability via scalable inline processing. Using a mathematical model for processing time (the proposed figure of merit), we present results which show that mmMoReEdge is 50% to 70% faster as compared to an offline general-purpose processor based architecture and is 30% to 40% faster as compared to a node-based architecture with one FPGA.

There is recent interest in using deep neural networks (DNNs) for controlling autonomous cyber-physical systems (CPSs). One challenge with this approach is that many autonomous CPS applications are safety-critical, and is not clear if DNNs can proffer safe system behaviors. To address this problem, we present an approach to modify existing (deep) reinforcement learning algorithms to guide the training of those controllers so that the overall system is safe. We present a novel verification-in-the-loop training algorithm that uses the formalism of barrier certificates to synthesize DNN-controllers that are safe by design. We demonstrate a proof-of-concept evaluation of our technique on multiple CPS examples.

Cloud computing is the newly emerged technology adopted by many organizations due to its different benefits. Unfortunately, despite all the benefits offered by the cloud, there are certain concerns regarding the security issues related to the cloud platform which can threaten its widespread adoption. In this study, we suggest a scalable model to profile and prevent security attacks in the application layer of a cloud environment using an accurate and interpretable machine learning algorithm called regularized class association rules. The proposed model is based, first, on three additional security factors $(k, l$ and $m)$ , second, on the traditional authentication methods such as passwords and biometrics including keystroke to grant access to the cloud services/resources for an authorized user. Moreover, a case study of the proposal is given in order to validate the model and its usefulness. Eventually, a simulation was done to test the model performances.

The conventional approach to pre-process data for compression is to apply transforms such as the Fourier, the Karhunen-Loeve, or wavelet transforms. One drawback from adopting such an approach is that it is independent of the use of the compressed data, which may induce significant optimality losses when measured in terms of final utility (instead of being measured in terms of distortion). We therefore revisit this paradigm by tayloring the data pre-processing operation to the utility function of the decision-making entity using the compressed (and therefore noisy) data. More specifically, the utility function consists of an Lp-norm, which is very relevant in the area of smart grids. Both a linear and a non-linear use-oriented transforms are designed and compared with conventional data pre-processing techniques, showing that the impact of compression noise can be significantlv reduced.

Augmented reality (AR) can support a user in performing an expert task by overlaying real world objects with the domain specific information required to complete the task. Understanding how users can process and use such information is very important for informing the design of AR technologies and applications. In this paper, the results of a quality of experience (QoE) evaluation of an AR application for the task of solving a Rubik's Cube are presented. The Rubik's Cube was selected based on its familiarity and the expertise needed to solve it unaided. An empirical approach was taken to identify the QoE features that affect the usability and utility of an AR head-mounted display (HMD) compared with paper-based instruction. The QoE evaluation methodology involved the capture and analysis of implicit and explicit QoE metrics. The utility (in terms of performance) of each mode of instruction was objectively measured using: (a) cube completion success rates; and (b) time-to-completion. The implicit metrics of electrodermal activity (EDA), skin temperature, heart rate and the novel use of facial action units (AUs) were recorded to infer emotional state during the task completion. Finally, with respect to explicit metrics, the test subjects completed a Likert scale questionnaire post the experience to subjectively report QoE as well as a self-assessment manikin (SAM) questionnaire to self-report emotional state upon task completion. The results show that AR yielded higher success rates and significantly lower time-to-completion rates. The AR group explicitly reported higher levels of positive valance (affective state) than the paper-based group. The physiological data showed that the AR group were less stressed (via EDA) than the paper-based group. Finally, analysis of the AU data reflected a greater than chance (total: 21.85%) accuracy when predicting affective state based on SAM questionnaires as ground-truth.

In this survey paper, we point out a comparative study on different types of plant leaf diseases and all the techniques of machine learning which had been used to detect plant leaf disease. This was accomplished for two principle reasons: to restrain the length of the paper and in light of the fact that techniques managing roots, seeds and natural products have a few idiosyncrasies that would warrant a particular study. The chose proposition are isolated into three classes as indicated by their goal: recognition, seriousness measurement, and characterization. Every one of those classes, thus, are subdivided by the primary specialized arrangement utilized in the calculation. This paper is required to be helpful to specialists working both on vegetable pathology and example acknowledgment, giving a thorough and open outline of this significant field of research.

This paper is a subset of a full-length paper that will be published in a special issue on metamaterials of the transactions at the end of this year. It presents the current authors’ perspective on the emerging field of spacetime metamaterials. Emphasis is placed on the definition, contextualization and fundamentals of the concept. In addition, a few new physical effects are described and a number of applications are enumerated.