Search

The integration of machine/deep learning and sensing technologies is transforming healthcare and medical practice. However, inherent limitations in healthcare data, namely scarcity, quality, and heterogeneity, hinder the effectiveness of supervised learning techniques which are mainly based on pure statistical fitting between data and labels. In this article, we first identify the challenges present in machine learning for pervasive healthcare and we then review the current trends beyond fully supervised learning that are developed to address these three issues. Rooted in the inherent drawbacks of empirical risk minimization that underpins pure fully supervised learning, this survey summarizes seven key lines of learning strategies, to promote the generalization performance for real-world deployment. In addition, we point out several directions that are emerging and promising in this area, to develop data-efficient, scalable, and trustworthy computational models, and to leverage multi-modality and multi-source sensing informatics, for pervasive healthcare. [ABSTRACT FROM AUTHOR]

Current-driven spin-orbit-torque (SOT)-induced device is a promising candidate with nonvolatility, low energy consumption, and ultrafast speed for the next-generation storage and computing technique. However, the requirement of the assistant magnetic field hinders its application. Besides, the switching current density in SOT-induced devices still needs to be further reduced. Here, we prepared devices with stacks of Ta/Pt100-xWx/Co/AlOy/Pt and systematically investigated changes in the switching efficiency with W content. A high damping-like effective field per unit current density ηDL up to 40.57 ± 3.32 (Oe/(106 A/cm2)) was observed in the device with a Pt74W26 layer, which is one order of magnitude higher than that in the typical spin-orbit devices with pure heavy metal layer reported in the previous articles. In addition, field-free switching is observed in devices with a wide range of W content using the competing spin currents generated from the Pt100-xWx layer. Zero-field switching (ZFS) critical current densities of them are less than 1.09 ± 0.05 (107 A/cm2) with the minimum of 1.58 ± 0.13 (106 A/cm2), indicating the highly efficient field-free spin-orbit switching in the PtW system. Our findings pave the way to high-energy-efficiency spin-orbit devices. [ABSTRACT FROM AUTHOR]

Autonomous underwater vehicles (AUVs) are commonly used to conduct complex underwater tasks, such as marine infrastructure overhaul and maintenance, environmental monitoring, oceanographic mapping, and organism capture. These tasks require the ability of an AUV to perform autonomous navigation, especially when communication is limited in the underwater environment. This brief developed a new type of lightweight intervention AUV for autonomous navigation using data from multiple inertial sensors, where multi-sensor error state Kalman filter schemes are preferable to standard Kalman Filters in terms of the AUV’s motion estimates. Concerning target recognition, a color restoration method is provided for degraded underwater images and a You Only Look Once strategy is combined with topological analysis for object detection. In addition, the proposed design is robust in terms of its software components and mechanical structure, which provides a feasible platform for AUV’s secondary development. Experiments of surveying and object manipulation conducted in underwater environments demonstrate the functionality of the entire system and its potential applications in the fields of science and industry. [ABSTRACT FROM AUTHOR]

In order to break limited step-down ratio range and poor regulation capacity in the conventional resonant switched-capacitor converters for rectifier application, a single-stage ac/dc rectifier based on the dual-resonance switched-capacitor converter is proposed in this article. Through utilizing a dual-resonant structure and nonoverlapping strategy to realize wide continuous voltage gain of the proposed converter, high power factor (PF) and low total harmonic distortion (THD) based on the narrow dead zone are guaranteed by operating at high step-down conversion ratio in this converter topology. Moreover, the voltage gain is concentrated between resonant frequency and double resonant frequency to achieve a narrow switching frequency range for rectifier application. Furthermore, operating at flat voltage stress curves to obtain low voltage stress character. Comprehensive mode analysis, input PF and THD, and stress analysis are given, and analysis results indicate that the proposed power factor correction converter can work with soft turn on and zero-current turn off; high efficiency is realized. Finally, an experimental prototype with 85–130 Vac input and 47 Vdc/44 W output is built and tested for verifying the operation modes and PFC performance, and the comparison between different rectifiers and the proposed one is provided. [ABSTRACT FROM AUTHOR]

Conventional biometric modalities, such as the face, fingerprint, and iris, are vulnerable against imitation and circumvention. Accordingly, secure biometric modalities with cancelable properties are needed for personal identification, especially in smart healthcare applications. Here we developed a person identification model using high-density surface electromyography (HD-sEMG) as biometric traits. In this model, the HD-sEMG biometric templates are cancelable and could be customized by the users through finger isometric contractions. A deep feature learning approach, implemented by convolutional neural networks (CNNs) is used to capture user-specific patterns from HD-sEMG signals and make identification decisions. This model has been validated on twenty-two subjects, with training and testing data acquired from two different days. The rank-1 identification accuracy and equal error rate for 44 identities (22 subjects × 2 accounts) can reach 87.23% and 4.66%, respectively. The cross-day identification accuracy of the proposed model is higher than the results of previous methods reported in the literature. The usability and efficiency of the proposed model are also investigated, indicating its potentials for practical applications. [ABSTRACT FROM AUTHOR]

In multiagent reinforcement learning (MARL), it is crucial for each agent to model the relation with its neighbors. Existing approaches usually resort to concatenate the features of multiple neighbors, fixing the size and the identity of the inputs. But these settings are inflexible and unscalable. In this article, we propose an attentive relational encoder (ARE), which is a novel scalable feedforward neural module, to attentionally aggregate an arbitrary-sized neighboring feature set for state representation in the decentralized MARL. The ARE actively selects the relevant information from the neighboring agents and is permutation invariant, computationally efficient, and flexible to interactive multiagent systems. Our method consistently outperforms the latest competing decentralized MARL methods in several multiagent tasks. In particular, it shows strong cooperative performance in challenging StarCraft micromanagement tasks and achieves over a 96% winning rate against the most difficult noncheating built-in artificial intelligence bots. [ABSTRACT FROM AUTHOR]