Northwestern Polytechnical University

Stochastic multiarmed bandits (stochastic MABs) are a problem of sequential decision-making with noisy rewards, where an agent sequentially chooses actions under unknown reward distributions to minimize cumulative regret. The majority of prior works on stochastic MABs assume that the reward distribution of each action has bounded supports or follows light-tailed distribution, i.e., sub-Gaussian distribution. However, in a variety of decision-making problems, the reward distributions follow a heavy-tailed distribution. In this regard, we consider stochastic MABs with heavy-tailed rewards, whose $p$ th moment is bounded by a constant $\nu_{p}$ for $1<p\leq2$ . First, we provide theoretical analysis on sub-optimality of the existing exploration methods for heavy-tailed rewards where it has been proven that existing exploration methods do not guarantee a minimax optimal regret bound. Second, to achieve the minimax optimality under heavy-tailed rewards, we propose a minimax optimal robust upper confidence bound (MR-UCB) by providing tight confidence bound of a $p$ -robust estimator. Furthermore, we also propose a minimax optimal robust adaptively perturbed exploration (MR-APE) which is a randomized version of MR-UCB. In particular, unlike the existing robust exploration methods, both proposed methods have no dependence on $\nu_{p}$ . Third, we provide the gap-dependent and independent regret bounds of proposed methods and prove that both methods guarantee the minimax optimal regret bound for a heavy-tailed stochastic MAB problem. The proposed methods are the first algorithm that theoretically guarantees the minimax optimality under heavy-tailed reward settings to the best of our knowledge. Finally, we demonstrate the superiority of the proposed methods in simulation with Pareto and Fréchet noises with respect to regrets.

This paper presents a multi-gas sensor device whose structure is optimized for edge computing capability under internet of things (IoT) environments. Considering inherent sensor device characteristics susceptible to environmental factors like temperature and humidity, edge-computing capability for the on-site sensor calibration and pattern recognition (PR) is facilitated through a proposed analog-assisted continual learning scheme. An environment-adaptable continual learning (EACL) is proposed to combine multiple learning processes under different environments including chamber and on-site. Its computation burden is much relieved to be integrated into the edge device by adopting the analog-assisted structure, where a designed readout integrated circuit (ROIC) for automatic calibration normalizes gas-sensor data. For functional feasibility, an edge-computing IoT device prototype is manufactured with a fabricated ROIC and an in-house semiconductor-type sensor array, supporting wireless on-site monitoring platform interfaces. The environment-adaptable edge-computing capability is functionally verified through EACL-PR experiments on hazardous gases such as NO 2 and CO under environmental factor variations. The average PR accuracy of 97% is achieved on several kinds of mixture gas patterns. The analog-assisted operation is verified to reduce the training cycles by 3 times while the EACL itself achieves 25% better efficiency.

This paper presents a robust predictor-based sliding mode control (RPSMC) for multirotor unmanned aerial vehicles (UAVs) to ensure desired tracking control under time delays which appear in practice by communications, complex computation, and actuator delays. Many UAV applications have difficulty in control and operation due to various types of time delays, resulting in repeated commands, severe control instability, and then mission failure. However, existing controllers have limitations in solving time delay problems. In this paper, the RPSMC with the prediction of future disturbance and reference trajectory is developed for the multirotor UAV to minimize the effects of time delay, robustly deal with external disturbances, and further achieve desired tracking control. The performance of RPSMC for the multirotor UAV is verified under various time delays and disturbances in numerical simulations. The results show the robustness and fast control convergence compared with proportional-integral-differential and conventional predictor-based controllers. Experimental results with step response and tracking of circular trajectory demonstrate the feasibility and performance of the RPSMC for UAVs in the presence of various time delays and disturbances.

Today's Internet must support applications with increasingly dynamic and heterogeneous connectivity requirements, such as video streaming and the Internet of Things. Yet current network management practices generally rely on pre-specified network configurations, which may not be able to cope with dynamic application needs. Moreover, even the best-specified policies will find it difficult to cover all possible scenarios, given applications' increasing heterogeneity and dynamic network conditions, e.g., on volatile wireless links. In this work, we instead propose a model-free learning approach to find the optimal network policies for current network flow requirements. This approach is attractive as comprehensive models do not exist for how different policy choices affect flow performance under changing network conditions. However, it can raise new challenges for online learning algorithms: policy configurations can affect the performance of multiple flows sharing the same network resources, and this performance coupling limits the scalability and optimality of existing online learning algorithms. In this work, we extend multi-armed bandit frameworks to propose new online learning algorithms for protocol selection with provably sublinear regret under certain conditions. We validate the optimality and scalability of our algorithms through data-driven simulations and testbed experiments.

The energy harvesting interface circuit is proposed for wind-driven triboelectric nanogenerator (WD-TENG). To extract power from the WD-TENG maximally and deliver power to the output (battery) efficiently, a rectifier-reusing bias-flip (RRBF) technique and a multi-phase reconfigurable switched-capacitor converter (MRSCC) are developed. In the RRBF, the low-side switches of the rectifier are reused as switches for bias-flip without additional components. The MRSCC delivers power to the battery efficiently by reducing switching loss including overlap loss. Furthermore, the MRSCC maintains a rectified voltage ( $V_{RECT}$ ) as high as a breakdown voltage ( $V_{BR}$ ) of a switch by adaptive conversion control and multi-phase operation to extract maximized power from the WD-TENG in a given process, even if the battery voltage is varied from 2.7 V to 4.2 V. Owing to the proposed techniques, the maximum extracted power to the output is 238 μW and peak power delivering efficiency of the MRSCC is 79.3%. The chip was fabricated in 0.18 μm BCD process.

A boost power factor correction (PFC) circuit has replaced the diode rectifier to improve its poor power factor performance, low efficiency, and output power limitation for conventional induction heating (IH) applications. Accordingly, many studies have been conducted, but they considered only the efficiency of the boost PFC rather than the entire IH system, or their control and design were complicated. In this paper, an algorithm tracking the maximum voltage gain of the resonant network is proposed to improve the entire efficiency of the two-stage IH system based on an exact online resonant frequency estimation. It can make the resonant network operate at the maximum voltage gain point which can improve the efficiency of the series-resonant inverter (SRI) included in the IH system with low circulating current, the minimum switching frequency, and zero voltage switching (ZVS) capability. The proposed algorithm also induces the minimum output voltage of the boost PFC, which can reduce its switching losses and total harmonic distortion (THD). The validity of the proposed algorithm is experimentally verified using a 2.4-kW prototype IH system, including the boost PFC and the IH-SRI controlled by a digital signal processor (DSP).

Although Resistive RAMs can support highly efficient matrix-vector multiplication, which is very useful for machine learning and other applications, the non-ideal behavior of hardware such as stuck-at fault and IR drop is an important concern in making ReRAM crossbar array-based deep learning accelerators. Previous work has addressed the nonideality problem through either redundancy in hardware, which requires a permanent increase of hardware cost, or software retraining, which may be even more costly or unacceptable due to its need for a training dataset as well as high computation overhead. In this paper we propose a very light-weight method that can be applied on top of existing hardware or software solutions. Our method, called FPT (Forward-Parameter Tuning), takes advantage of a certain statistical property existing in the activation data of neural network layers, and can mitigate the impact of mild nonidealities in ReRAM crossbar arrays for deep learning applications without using any hardware, a dataset, or gradientbased training. Our experimental results using MNIST, CIFAR-10, CIFAR-100, and ImageNet datasets in binary and multibit networks demonstrate that our technique is very effective, both alone and together with previous methods, up to 20rate, which is higher than even some of the previous remapping methods. We also evaluate our method in the presence of other nonidealities such as variability and IR drop. Further, we provide an analysis based on the concept of effective fault rate, which not only demonstrates that effective fault rate can be a useful tool to predict the accuracy of faulty RCA-based neural networks, but also explains why mitigating the SAF problem is more difficult with multi-bit neural networks.

An impedance plethysmogram (IPG) IC with wide-range fine baseline tracking capability is proposed to provide continuous blood pressure (BP) and cardiovascular monitoring just by wearing a neckband wearable device. For the baseline tracking, a mixed-mode baseline cancellation (MM-BC) scheme is proposed to achieve wide dynamic range (DR) and high SNR performance, where an artifact-detecting continuous-time (CT)- $\Delta \Sigma $ ADC is included together for high-resolution conversion. For cost-effective IPG design, a phase synchronizer concept is introduced to provide I/Q detection capability with single readout channel, supporting both configurations of two-electrode and four-electrode. A proposed readout integrated circuit (ROIC) prototype for the proposed IPG was fabricated, where an electrocardiogram (ECG) readout path is integrated together. The proposed IPG schemes were experimentally verified to achieve good performances of 145.2-dB DR and 103.5-dB SNR. For system-level feasibility, a neckband device prototype was also manufactured, and its cardiovascular monitoring functionality was functionally verified. Based on its IPG and ECG measurements, continuous BP measurement based on the pulse arrival time (PAT) was experimentally verified with a reference BP device.

We present Roslingifier, a data-driven storytelling method for animated scatterplots. Like its namesake, Hans Rosling (1948--2017), a professor of public health and a spellbinding public speaker, Roslingifier turns a sequence of entities changing over time---such as countries and continents with their demographic data---into an engaging narrative telling the story of the data. This data-driven storytelling method with an in-person presenter is a new genre of storytelling technique and has never been studied before. In this paper, we aim to define a design space for this new genre---data presentation---and provide a semi-automated authoring tool for helping presenters create quality presentations. From an in-depth analysis of video clips of presentations using interactive visualizations, we derive three specific techniques to achieve this: natural language narratives, visual effects that highlight events, and temporal branching that changes playback time of the animation. Our implementation of the Roslingifier method is capable of identifying and clustering significant movements, automatically generating visual highlighting and a narrative for playback, and enabling the user to customize. From two user studies, we show that Roslingifier allows users to effectively create engaging data stories and the system features help both presenters and viewers find diverse insights.

In this paper, we present a new transportation mode recognition system for smartphones called DeepVehicleSense, which is widely applicable to mobile context-aware services. DeepVehicleSense aims at achieving three performance objectives: high accuracy, low latency, and low power consumption at once by exploiting sound characteristics captured from the built-in microphone while being on candidate transportations. To attain high energy efficiency, DeepVehicleSense adopts hierarchical accelerometer-based triggers that minimize the activation of the microphone of smartphones. Further, to achieve high accuracy and low latency, DeepVehicleSense makes use of non-linear filters that can best extract the transportation sound samples. For recognition of five different transportation modes, we design a deep learning based sound classifier using a novel deep neural network architecture with multiple branches. Our staged inference technique can significantly reduce runtime and energy consumption while maintaining high accuracy for the majority of samples. Through 263-hour datasets collected by seven different Android phone models, we demonstrate that DeepVehicleSense achieves the recognition accuracy of 97.44\% with only sound samples of 2 seconds at the power consumption of 35.08 mW on average for all-day monitoring.

This article proposes a distributed estimation algorithm that uses local information about the neighbors through sensing or communication to design an estimation-based cooperative control of the stochastic multiagent system (MAS). The proposed distributed estimation algorithm solely relies on local sensing information rather than exchanging estimated state information from other agents, as is commonly required in conventional distributed estimation methods, reducing communication overhead. Furthermore, the proposed method allows interactions between all agents, including non-neighboring agents, by establishing a virtual fully connected network with the MAS state information independently estimated by each agent. The stability of the proposed distributed estimation algorithm is theoretically verified. Numerical simulations demonstrate the enhanced performance of the estimation-based linear and nonlinear control. In particular, using the virtual fully connected network concept in the MAS with the sensing/communication range, the flock configuration can be tightly controlled within the desired boundary, which cannot be achieved through the conventional flocking methods.

This article conducts a comparative analysis to investigate the effects of different classification algorithms and structural proximity indexes on the performance of the supervised link prediction approach to anticipating technology convergence at different forecast horizons. For this, we identify relationships between technologies of interest for different time periods and compute 10 structural proximity indexes among unconnected technologies at each period. We develop a set of classification models that identify potential convergence among unconnected technologies where each model is configured differently by a classification algorithm and a combination of the proximity indexes. We compare the performance of the classification models to investigate effective combinations of classification algorithms and proximity indexes at different forecast horizons. The empirical analysis on Wikipedia articles about artificial intelligence technology indicates that random forest outperforms others in short-term forecasting while support vector machine outperforms others in mid-term forecasting. We also identify structural proximity indexes that produce higher performance when combined with the most effective algorithm at each forecast horizon. The results of this article are expected to offer guidelines for choosing classification algorithms and indexes when applying the supervised link prediction approach in anticipating technology convergence.

With advances in flexible electronics, innovative foldable, rollable, and stretchable displays have been developed to maintain their performance under various deformations. These flexible devices can develop more innovative designs than conventional devices due to their light weight, high space efficiency, and practical convenience. However, developing flexible devices requires material innovation because the devices must be flexible and exhibit desirable electrical insulating/semiconducting/metallic properties. Recently, emerging two-dimensional materials such as graphene, hexagonal boron nitride, and transition metal dichalcogenides have attracted considerable research attention because of their outstanding electrical, optical, and mechanical properties, which are ideal for flexible electronics. We review the recent progress and challenges of two-dimensional material growth and display applications, and discuss perspectives for exploring two-dimensional materials for display applications. This article is protected by copyright. All rights reserved.

Secure precoding superimposed with artificial noise (AN) is a promising transmission technique to improve security by harnessing the superposition nature of the wireless medium. However, finding a jointly optimal precoding and AN structure is very challenging in downlink multi-user multiple-input multiple-output wiretap channels with multiple eavesdroppers. The major challenge in maximizing the secrecy rate arises from the non-convexity and non-smoothness of the rate function. Traditionally, an alternating optimization framework that identifies beamforming vectors and AN covariance matrix has been adopted; yet this alternating approach has limitations in maximizing the secrecy rate. In this paper, we put forth a novel secure precoding algorithm that jointly and simultaneously optimizes the beams and AN covariance matrix for maximizing the secrecy rate when a transmitter has either perfect or partial channel knowledge of eavesdroppers. To this end, we first establish an approximate secrecy rate in a smooth function. Then, we derive the first-order optimality condition in the form of the nonlinear eigenvalue problem (NEP). We present a computationally efficient algorithm to identify the principal eigenvector of the NEP as a suboptimal solution for secure precoding. Simulations demonstrate that the proposed methods improve secrecy rate significantly compared to the existing methods.

Objective : Continuous glucose monitoring system (CGMS) is growing popular and preferred by diabetes over conventional methods of self-blood glucose monitoring (SBGM) systems. However, currently available commercial CGMS in the market is useful for few days to few months. This paper presents a durable, highly sensitive and minimally invasive implant type electromagnetic sensor for continuous glucose monitoring that is capable of tracking minute changes in blood glucose level (BGL). Methods : The proposed sensor utilizes strong oscillating nearfield to detect minute changes in dielectric permittivity of interstitial fluid (ISF) and blood due to changes in BGL. A biocompatible packaging material is used to cover the sensor. It helps in minimizing foreign body reactions (FBR) and improves stability of the sensor. Results : The performance of the proposed sensor was evaluated on live rodent models (C57BL/6J mouse and Sprague Dawley rat) through intravenous glucose and insulin tolerance tests. Biocompatible polyolefin was used as the sensor packaging material, and the effect of packaging thickness on the sensitivity of sensor was examined in in-vivo test. Proposed sensor could track real-time BGL change measured with a commercial blood glucose meter. High linear correlation (R 2 > 0.9) with measured BGL was observed during in vivo experiments. Conclusion : The experimental results demonstrate that the proposed sensor is suitable for long term CGMS applications with a high accuracy. Significance : Present work offers a new perspective towards development of long term CGM system using electromagnetic based implant sensor. The in vivo evaluation of the sensor shows excellent tracking of BGL changes.