Asymmetric Network Research Articles

SiamPolar: Semi-supervised realtime video object segmentation with polar representation

Video object segmentation (VOS) is an essential part of autonomous vehicle navigation. Besides the accuracy metric, the real-time speed is very important for the algorithms of autonomous vehicles. In this paper, we propose a semi-supervised real-time method based on the Siamese network using a new polar representation. The input of bounding boxes are initialized rather than the object masks, which are applied to the video object detection tasks. The polar representation could reduce the parameters for encoding masks with subtle accuracy loss, so that the algorithm speed can be improved significantly. An asymmetric siamese network is also developed to extract the features from different spatial scales. Moreover, the peeling convolution is proposed to reduce the antagonism among the branches of the polar head. The repeated cross-correlation and semi-FPN are designed based on this idea. The experimental results on the DAVIS-2016 dataset and other public datasets demonstrate the effectiveness of the proposed method.

Series AC Arc Fault Diagnosis Based on Data Enhancement and Adaptive Asymmetric Convolutional Neural Network

Arc fault diagnosis methods based on traditional machine learning rely heavily on expert-designed characteristics and are difficult to train models with strong generalization. Deep learning arc fault detection method requires a large number of experimental data to meet the diagnostic performance, and part of the experimental data is difficult to obtain, which increases the difficulty of model training. This paper presents an adaptive arc fault diagnosis model which provides data enhancement. In the data processing stage, the continuous time series can be discretely divided according to the half-period length, and the adaptive matrix transformation of the discrete sequence is carried out according to the convolution operation. At the same time, the generative adversarial network is used to enhance the original coding data of the carbonized path arc fault. In the training stage, an adaptive asymmetric convolution kernel is designed according to the data distribution characteristics of the coding matrix, which can effectively mine the continuous time information and correlation information in the time domain. Compared with the traditional convolutional neural network, the generative adversarial network can effectively enhance the data of carbonized path arc fault and improve the robustness of the diagnosis model. The adaptive asymmetric convolutional neural network improves the feature extraction of arc faults by convolutional layer and the ability of arc fault diagnosis.

Asymmetric Risk Spillover Networks and Risk Contagion Driver in Chinese Financial Markets: The Perspective of Economic Policy Uncertainty

The global financial market shocks have intensified due to the COVID-19 epidemic and other impacts, and the impacts of economic policy uncertainty on the financial system cannot be ignored. In this paper, we construct asymmetric risk spillover networks of Chinese financial markets based on five sectors: bank, securities, insurance, diversified finance, and real estate. We investigate the complexity of the risk spillover effect of Chinese financial markets and the impact of economic policy uncertainty on the level of network contagion of financial risk. The study yields three findings. First, the cross-sectoral risk spillover effects of Chinese financial markets are asymmetric in intensity. The bank sector is systemically important in the risk spillover network. Second, the level of risk stress in the real estate sector has increased in recent years, and it plays an important role in the path of financial risk contagion. Third, Economic policy uncertainty has a significant positive impact on the level of network contagion of financial risk of Chinese financial markets.

Flexible and Anisotropic Strain Sensors with the Asymmetrical Cross-Conducting Network for Versatile Bio-Mechanical Signal Recognition.

Flexible strain sensors with high performance are actively and widely investigated for wearable electronic devices. However, the conventional sensors often suffer from a lack of detection of complex multidimensional strain, which severely limits their wide applications. To overcome this critical challenge, we propose a pattern design by screen printing to construct an asymmetrical cross-conductive network in the piezoresistive strain sensor, which can enhance the response to external stimuli in different directions. The unique network endows the prepared sensors with the excellent ability of instantaneous detection and accurate identification of multidimensional strains. Moreover, the sensor also demonstrates high sensitivity, fast response, an ultra-wide sensing range, and excellent stability and durability. Benefiting from the outstanding comprehensive performance of the prepared sensor, a full range of human actions (wink, smile, swallowing, and joint bending) and subtle bio-signals (pulse and breathing) are easily and accurately monitored. A wireless wearable device assembled by the sensor shows great potential applications in practical real-time physiological monitoring and intelligent mobile diagnosis for humans. This work provides an innovative and effective strategy for manufacturing flexible and multifunctional strain sensors to fully satisfy versatile applications of new-generation wearable electronic devices.

A 0.35-V 5,200-μm2 2.1-MHz Temperature-Resilient Relaxation Oscillator With 667 fJ/Cycle Energy Efficiency Using an Asymmetric Swing-Boosted RC Network and a Dual-Path Comparator

This article describes a 2.1-MHz relaxation oscillator (RxO) for energy-harvesting Internet-of-Things (IoT) sensor nodes. The RxO features an asymmetric swing-boosted <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">RC</i> network and a dual-path comparator to surmount the challenges of sub-0.5-V operation while achieving temperature resilience. The former enables alternating the common-mode voltages at the output of the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">RC</i> network to facilitate the sub-0.5-V operation, while the latter is outfitted with a delay generator for tracking the temperature-sensitive delay of the comparator. Prototyped in 28-nm CMOS, the RxO occupies a tiny footprint of 5,200 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mu \text {m}^{2}$ </tex-math></inline-formula> . The power consumption is 1.4 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mu \text{W}$ </tex-math></inline-formula> at 0.35 V. The measured temperature stability is 158 ppm/°C (average of seven chips) over −20 °C–120 °C. It scores the best energy efficiency (667 fJ/cycle) among the reported MHz-range RxOs and has a figure-of-merit (181 dB) that compares favorably with the state-of-the-art.

Asymmetric coordinate attention spectral-spatial feature fusion network for hyperspectral image classification

In recent years, the hyperspectral classification algorithm based on deep learning has received widespread attention, but the existing network models have higher model complexity and require more time consumption. In order to further improve the accuracy of hyperspectral image classification and reduce model complexity, this paper proposes an asymmetric coordinate attention spectral-spatial feature fusion network (ACAS2F2N) to capture distinguishing hyperspectral features. Specifically, adaptive asymmetric iterative attention was proposed to obtain the discriminative spectral-spatial features. Different from the common feature fusion method, this feature fusion method can adapt to most skip connection tasks. In addition, there is no manual parameter setting. Coordinate attention is used to obtain accurate coordinate information and channel relationship. The strip pooling module was introduced to increase the network’s receptive field and avoid irrelevant information brought by conventional convolution kernels. The proposed algorithm is tested on the mainstream hyperspectral datasets (IP, KSC, and Botswana), experimental results show that the proposed ACAS2F2N can achieve state-of-the-art performance with lower time complexity.

J-Net: Asymmetric Encoder-Decoder for Medical Semantic Segmentation

With the development of deep learning, breakthroughs have been made in the field of semantic segmentation. However, it is difficult to generate a fine mask on the same medical images because medical images have low contrast, high resolution, and insufficient semantic information. In most scenarios, existing approaches mostly use a pooling layer to reduce the resolution of feature maps. Therefore, it is difficult for them to consider the whole image features, resulting in information loss and performance degradation. In this paper, a multiscale asymmetric encoder-decoder semantic segmentation network is proposed. The network consists of two parts, which perform feature extraction and image restoration on the input, respectively. The encoder network obtains multiscale feature information by connecting multiple ASPP modules to form a feature pyramid. Meanwhile, the upsampling layer of each decoder can be connected to the feature map generated by the corresponding ASPP module. Finally, the classification information of each pixel is obtained through the sigmoid function. The performance of the proposed method can be verified on publicly available datasets. The experimental evidence shows that the proposed method can take full advantage of multiscale feature information and achieve superior performance with less inference computational cost.

A Proportional-Egalitarian Allocation Policy for Public Goods Problems with Complex Network

How free-riding behavior can be avoided is a constant topic in public goods problems, especially in persistent and complex resource allocation situations. In this paper, a novel allocation policy for public goods games with a complex network, called the proportional-egalitarian allocation method (PEA), is proposed. This allocation rule differs from the well-studied redistribution policies by following a two-step process without paying back into the common pool. A parameter is set up for dividing the total income into two parts, and then they are distributed by following the egalitarianism and proportional rule, respectively. The first part of total income is distributed equally, while the second part is allocated proportionally according to players’ initial payoffs. In addition, a new strategy-updating mechanism is proposed by comparing the average group payoffs instead of the total payoffs. Compared with regular lattice networks, this mechanism admits the difference of cooperative abilities among players induced by the asymmetric network. Furthermore, numerical calculations show that a relatively small income for the first distribution step will promote the cooperative level, while relatively less income for the second step may harm cooperation evolution. This work thus enriches the knowledge of allocation policies for public goods games and also provides a fresh perspective for the strategy-updating mechanism.

Minimizing total interference in asymmetric sensor networks

The problem of computing a connected network with minimum interference is a fundamental problem in wireless sensor networks. Several models of interference have been studied in the literature. The most common one is the receiver-centric, in which the interference of a node p is defined as the number of other nodes whose transmission range covers p. In this paper, we study the problem of assigning a transmission range to each sensor, such that the resulting network is strongly connected and the total interference of the network is minimized.For the one-dimensional case, we show how to solve the problem optimally in O(n3) time. For the two-dimensional case, we show that the problem is NP-complete and give a polynomial-time 2-approximation algorithm for the problem.

Asymmetric multi-task attention network for prostate bed segmentation in computed tomography images.

Post-prostatectomy radiotherapy requires accurate annotation of the prostate bed (PB), i.e., the residual tissue after the operative removal of the prostate gland, to minimize side effects on surrounding organs-at-risk (OARs). However, PB segmentation in computed tomography (CT) images is a challenging task, even for experienced physicians. This is because PB is almost a "virtual" target with non-contrast boundaries and highly variable shapes depending on neighboring OARs. In this work, we propose an asymmetric multi-task attention network (AMTA-Net) for the concurrent segmentation of PB and surrounding OARs. Our AMTA-Net mimics experts in delineating the non-contrast PB by explicitly leveraging its critical dependency on the neighboring OARs (i.e., the bladder and rectum), which are relatively easy to distinguish in CT images. Specifically, we first adopt a U-Net as the backbone network for the low-level (or prerequisite) task of the OAR segmentation. Then, we build an attention sub-network upon the backbone U-Net with a series of cascaded attention modules, which can hierarchically transfer the OAR features and adaptively learn discriminative representations for the high-level (or primary) task of the PB segmentation. We comprehensively evaluate the proposed AMTA-Net on a clinical dataset composed of 186 CT images. According to the experimental results, our AMTA-Net significantly outperforms current clinical state-of-the-arts (i.e., atlas-based segmentation methods), indicating the value of our method in reducing time and labor in the clinical workflow. Our AMTA-Net also presents better performance than the technical state-of-the-arts (i.e., the deep learning-based segmentation methods), especially for the most indistinguishable and clinically critical part of the PB boundaries. Source code is released at https://github.com/superxuang/amta-net.

High efficiency Doherty power amplifier based on asymmetrical matching network

Doherty power amplifier (DPA) with high efficiency at the output power back off is highly demanded for modern wireless communication systems to achieve high data rates and reduce the power consumption and operation costs. This paper presents a new design strategy for enhancing DPA’s back-off efficiency. New design strategy called asymmetrical matching network is used to achieve asymmetric operation, which helps to compensate for the low power delivered by the peaking stage in the conventional DPA. The simulation results showed an enhancement in the back-off efficiency, where the proposed design is able toachieve 46-52% drain efficiency at 8 dB output power back-off while maintains high efficiency of 73-80 % at saturation over the designed bandwidth of 3.4-3.6 GHz. The proposed design is suitable for high efficiency sub-6 GHz fifth-generation wireless applications.&lt;br /&gt;&lt;div&gt; &lt;/div&gt;

ECCENTRIC DISTANCE SUM OF SUBSTITUTION TREE NETWORKS

In this paper, we study the eccentric distance sum of substitution tree networks. Calculation of eccentric distance sum naturally involves calculation of average geodesic distance and it is much more complicated. We obtain the asymptotic formulas of average geodesic distance and eccentric distance sum of both symmetric and asymmetric substitution tree networks. Our result on average geodesic distance generalizes the result of [T. Li, K. Jiang and L. Xi, Average distance of self-similar fractal trees, Fractals 26(1) (2018) 1850016.] from symmetric case to asymmetric case. To derive formulas, we investigate the corresponding integrals on self-similar measure and use the self-similarity of distance and measure. For simplicity, we introduce some systematic symbolic assignments and make some assumptions on the graph. We verify that our formulas are correct using the numerical calculation results.

Operating Cost Reduction in Distribution Networks Based on the Optimal Phase-Swapping including the Costs of the Working Groups and Energy Losses

The problem of optimal phase-balancing in three-phase asymmetric distribution networks is addressed in this research from the point of view of combinatorial optimization using a master–slave optimization approach. The master stage employs an improved sine cosine algorithm (ISCA), which is entrusted with determining the load reconfiguration at each node. The slave stage evaluates the energy losses for each set of load connections provided by the master stage by implementing the triangular-based power flow method. The mathematical model that was solved using the ISCA is designed to minimize the annual operating costs of the three-phase network. These costs include the annual costs of the energy losses, considering daily active and reactive power curves, as well as the costs of the working groups tasked with the implementation of the phase-balancing plan at each node. The peak load scenario was evaluated for a 15-bus test system to demonstrate the effectiveness of the proposed ISCA in reducing the power loss (18.66%) compared with optimization methods such as genetic algorithm (18.64%), the classical sine cosine algorithm (18.42%), black-hole optimizer (18.38%), and vortex search algorithm (18.59%). The IEEE 37-bus system was employed to determine the annual total costs of the network before and after implementing the phase-balancing plan provided by the proposed ISCA. The annual operative costs were reduced by about 13% with respect to the benchmark case, with investments between USD 2100 and USD 2200 in phase-balancing activities developed by the working groups. In addition, the positive effects of implementing the phase-balancing plan were evidenced in the voltage performance of the IEEE 37-bus system by improving the voltage regulation with a maximum of 4% in the whole network from an initial regulation of 6.30%. All numerical validations were performed in the MATLAB programming environment.

Identification and Diagnosis of Cerebral Stroke through Deep Convolutional Neural Network-Based Multimodal MRI Images.

This study aimed to explore the application value of multimodal magnetic resonance imaging (MRI) images based on the deep convolutional neural network (Conv.Net) in the diagnosis of strokes. Specifically, four automatic segmentation algorithms were proposed to segment multimodal MRI images of stroke patients. The segmentation effects were evaluated factoring into DICE, accuracy, sensitivity, and segmentation distance coefficient. It was found that although two-dimensional (2D) full convolutional neural network-based segmentation algorithm can locate and segment the lesion, its accuracy was low; the three-dimensional one exhibited higher accuracy, with various objective indicators improved, and the segmentation accuracy of the training set and the test set was 0.93 and 0.79, respectively, meeting the needs of automatic diagnosis. The asymmetric 3D residual U-Net network had good convergence and high segmentation accuracy, and the 3D deep residual network proposed on its basis had good segmentation coefficients, which can not only ensure segmentation accuracy but also avoid network degradation problems. In conclusion, the Conv.Net model can accurately segment the foci of patients with ischemic stroke and is suggested in clinic.

Micro-CT image denoising with an asymmetric perceptual convolutional network

Micro-CT has important applications in biomedical research due to its ability to perform high-precision 3D imaging of micro-architecture in a non-invasive way. Because of the limited power of the radiation source, it is difficult to obtain a high signal-to-noise image under the requirement of temporal resolution. Therefore, low-dose CT image denoising has attracted considerable attention to improve the image quality of micro-CT while maintaining time resolution. In this paper, an end-to-end asymmetric perceptual convolutional network (APCNet) is proposed to enhance the network’s ability to capture and retain image details by improving the convolutional layer and introducing an edge detection layer. Compared with the previously proposed denoising models such as DnCNN, CNN-VGG, and RED-CNN, experiments proved that our proposed method has achieved better results in both numerical indicators and visual perception.

Connectional asymmetry of the inferior parietal lobule shapes hemispheric specialization in humans, chimpanzees, and rhesus macaques.

The inferior parietal lobule (IPL) is one of the most expanded cortical regions in humans relative to other primates. It is also among the most structurally and functionally asymmetric regions in the human cerebral cortex. Whether the structural and connectional asymmetries of IPL subdivisions differ across primate species and how this relates to functional asymmetries remain unclear. We identified IPL subregions that exhibited positive allometric in both hemispheres, scaling across rhesus macaque monkeys, chimpanzees, and humans. The patterns of IPL subregions asymmetry were similar in chimpanzees and humans, but no IPL asymmetries were evident in macaques. Among the comparative sample of primates, humans showed the most widespread asymmetric connections in the frontal, parietal, and temporal cortices, constituting leftward asymmetric networks that may provide an anatomical basis for language and tool use. Unique human asymmetric connectivity between the IPL and primary motor cortex might be related to handedness. These findings suggest that structural and connectional asymmetries may underlie hemispheric specialization of the human brain.

A 24–44 GHz Broadband Transmit–Receive Front End in 0.13-μm SiGe BiCMOS for Multistandard 5G Applications

In this article, a novel broadband millimeter-wave (mm-wave) front end (FE) based on the asymmetric switch network structure is proposed for multistandard fifth-generation (5G) applications. The asymmetric switch network consists of a bidirectional impedance transformation network and a T-type topology single-pole single-throw (SPST) switch. In the TX-mode, the bidirectional impedance transformation network serves as an output matching network to achieve a continuous-mode Class-F <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-1</sup> power amplifier (PA) to expand the bandwidth. In the RX-mode, the OFF impedance of the PA can be transformed into a high-impedance region by the bidirectional impedance transformation network. Based on the proposed structure, a broadband mm-wave front end is implemented in the 0.13- μm SiGe BiCMOS process, which occupies 906 μm ×997 μm including pads. In the TX-mode, the -1-dB P <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">sat</sub> bandwidth of the FE covers 21-44 GHz with the maximum output power of 18.1-19 dBm and the peak power added efficiency (PAE) of 20.2%-26.8%. In the RX-mode, the FE achieves an 84.6% and -3-dB small-signal gain fractional bandwidth from 20.2 to 49.8 GHz with 38-mW power consumption. It achieves a measured noise figure (NF) of 7.62-11.64 dB over the frequency range of 24-48 GHz. To the best of our knowledge, this is the first FE that covers all the present 5G mm-wave bands without reconfiguration.

LAENet: Light-weight asymmetric encoder-decoder network for semantic segmentation

Encode-decoder structure is used in deep learning for real-time dense segmentation task. On account of the limitation of calculation burden on mobile devices, we present a light-weight asymmetric encoder-decoder network in this paper, namely LAENet, which quickly and efficiently accomplish the task of real-time semantic segmentation. We employ an asymmetric convolution and group convolution structure combined with dilated convolution and dense connectivity to reduce computation cost and model size, which can guarantee adequate receptive field and enhance the model learning ability in encoder. On the other hand, feature pyramid networks (FPN) structure combine attention mechanism and ECRE block are utilized in the decoder to strike a balance between the network complexity and segmentation performance. Our approach achieves only have 0.84M parameters, and is able to reach 66 FPS in a single GTX 1080Ti GPU. Experiments on Cityscapes datasets demonstrate that superior performance of LAENet is better than the existing segmentation network, in terms of speed and accuracy trade-off without any post-processing.

Ionic Diode Based on an Asymmetric‐Shaped Carbon Black Nanoparticle Membrane

AbstractBioinspired ionic diodes are widely explored to mimic the controllable ion transport of biological ion channels. However, due to their vertical structures, the integration of conventional ionic diodes into complex ionic circuits is still a challenge. Here, a horizontal ionic diode is developed based on an asymmetric nanochannel network membrane (NCNM) constructed from carbon black nanoparticles. The rectification of ionic current is achieved through the asymmetric concentration polarization of ions at two ends of the asymmetric NCNM. The rectification ratio of the NCNM ionic diode can be modified flexibly by changing the working fluid and the geometry of the NCNM. It is found that with the presence of cationic surfactant in the working fluid, the rectification ratio increases more than 30 times from 3.03 to 109.77. Advanced functions of the developed ionic component, including working as an ionic transistor for current switching and integrating into an ionic diode bridge on a single nanofluidic chip for rectifying alternating current signals, are also demonstrated in this paper. The horizontally arranged NCNM ionic diode possesses the advantages of easy fabrication and integration that can be practically applied in the development of ionic electronics and biocomputing.

Integrated Inference of Asymmetric Protein Interaction Networks Using Dynamic Model and Individual Patient Proteomics Data

Recent advances in experimental biology studies have produced large amount of molecular activity data. In particular, individual patient data provide non-time series information for the molecular activities in disease conditions. The challenge is how to design effective algorithms to infer regulatory networks using the individual patient datasets and consequently address the issue of network symmetry. This work is aimed at developing an efficient pipeline to reverse-engineer regulatory networks based on the individual patient proteomic data. The first step uses the SCOUT algorithm to infer the pseudo-time trajectory of individual patients. Then the path-consistent method with part mutual information is used to construct a static network that contains the potential protein interactions. To address the issue of network symmetry in terms of undirected symmetric network, a dynamic model of ordinary differential equations is used to further remove false interactions to derive asymmetric networks. In this work a dataset from triple-negative breast cancer patients is used to develop a protein-protein interaction network with 15 proteins.