arXiv subjects

Yichen Zhang

Publications and source records attributed to Yichen Zhang.

At least 199 records · Page 11Linked to original sources

arXiv2020

Adaptive Switching Control of Wind Turbine Generators for Necessary Frequency Response

This letter proposes a new control strategy for wind turbine generators to decide the necessity of switches between the normal operation and frequency support modes. The idea is to accurately predict an unsafe frequency response using a differential transformation method right after power imbalance is detected so as to adaptively activate a frequency support mode only when necessary. This control strategy can effectively avoid unnecessary switches with a conventional use of deadband but still ensure adequate frequency response.

eess.SY↗

arXiv2020

The High-Mass Protostellar Population of a Massive Infrared Dark Cloud

We conduct a census of the high-mass protostellar population of the $\sim70,000\:M_\odot$ Infrared Dark Cloud (IRDC) G028.37+00.07, identifying 35 sources based on their $70\:μ$m emission, as reported in the {\it Herschel} Hi-GAL catalog of Molinari et al. (2016). We perform aperture photometry to construct spectral energy distributions (SEDs), which are then fit with the massive protostar models of Zhang & Tan (2018). We find that the sources span a range of isotropic luminosities from $\sim$20 to 4,500$\:L_\odot$. The most luminous sources are predicted to have current protostellar masses of $m_{*}\sim10\:M_\odot$ forming from cores of mass $M_{c}\sim40$ to $400\:M_\odot$. The least luminous sources in our sample are predicted to be protostars with masses as low as $\sim 0.5\:M_\odot$ forming from cores with $M_{c}\sim10\:M_\odot$, which are the minimum values explored in the protostellar model grid. The detected protostellar population has a total estimated protostellar mass of $M_{*}\sim 100\:M_\odot$. Allowing for completeness corrections, which are constrained by comparison with an ALMA study in part of the cloud, we estimate a star formation efficiency per free-fall time of $\sim3\%$ in the IRDC. Finally, analyzing the spatial distribution of the sources, we find relatively low degrees of central concentration of the protostars. The protostars, including the most massive ones, do not appear to be especially centrally concentrated in the protocluster as defined by the IRDC boundary.

astro-ph.GA↗

arXiv2020

On the Modeling and Simulation of Anti-Windup Proportional-Integral Controller

This paper investigates the chattering and deadlock behaviors of the proportional-integral (PI) controller with an anti-windup (AW) limiter recommended by the IEEE Standard 421.5-2016. Depending on the simulation method, the controller may enter a chattering or deadlock state in some combinations of parameters and inputs. Chattering and deadlock are analyzed in the context of three numerical integration approaches: explicit partitioned method (EPM), execution-list based method (ELM), and implicit trapezoidal method (ITM). This paper derives the chattering stop condition for EPM and ELP, and analyzes the impacts of step size and convergence tolerance for simultaneous method. The deduced chattering stop conditions and deadlock behavior is verified with numerical simulations.

eess.SY↗

arXiv2020

Substructure Formation in a Protostellar Disk of L1527 IRS

We analyze multi-frequency, high-resolution continuum data obtained by ALMA and JVLA to study detailed structure of the dust distribution in the infant disk of a Class~0/I source, L1527 IRS. We find three clumps aligning in the north-south direction in the $7 {\rm \,mm}$ radio continuum image. The three clumps remain even after subtracting free-free contamination, which is estimated from the $1.3{\rm \,cm}$ continuum observations. The northern and southern clumps are located at a distance of $\sim 15{\rm \,au}$ from the central clump and are likely optically thick at $7{\rm \,mm}$ wavelength. The clumps have similar integrated intensities. The symmetric physical properties could be realized when a dust ring or spiral arms around the central protostar is projected to the plane of the sky. We demonstrates for the first time that such substructure may form even in the disk-forming stage, where the surrounding materials actively accrete toward a disk-protostar system.

astro-ph.SR↗

arXiv2020

HRank: Filter Pruning using High-Rank Feature Map

Neural network pruning offers a promising prospect to facilitate deploying deep neural networks on resource-limited devices. However, existing methods are still challenged by the training inefficiency and labor cost in pruning designs, due to missing theoretical guidance of non-salient network components. In this paper, we propose a novel filter pruning method by exploring the High Rank of feature maps (HRank). Our HRank is inspired by the discovery that the average rank of multiple feature maps generated by a single filter is always the same, regardless of the number of image batches CNNs receive. Based on HRank, we develop a method that is mathematically formulated to prune filters with low-rank feature maps. The principle behind our pruning is that low-rank feature maps contain less information, and thus pruned results can be easily reproduced. Besides, we experimentally show that weights with high-rank feature maps contain more important information, such that even when a portion is not updated, very little damage would be done to the model performance. Without introducing any additional constraints, HRank leads to significant improvements over the state-of-the-arts in terms of FLOPs and parameters reduction, with similar accuracies. For example, with ResNet-110, we achieve a 58.2%-FLOPs reduction by removing 59.2% of the parameters, with only a small loss of 0.14% in top-1 accuracy on CIFAR-10. With Res-50, we achieve a 43.8%-FLOPs reduction by removing 36.7% of the parameters, with only a loss of 1.17% in the top-1 accuracy on ImageNet. The codes can be available at https://github.com/lmbxmu/HRank.

cs.CV↗

arXiv2020

Decentralized Visual-Inertial-UWB Fusion for Relative State Estimation of Aerial Swarm

The collaboration of unmanned aerial vehicles (UAVs) has become a popular research topic for its practicability in multiple scenarios. The collaboration of multiple UAVs, which is also known as aerial swarm is a highly complex system, which still lacks a state-of-art decentralized relative state estimation method. In this paper, we present a novel fully decentralized visual-inertial-UWB fusion framework for relative state estimation and demonstrate the practicability by performing extensive aerial swarm flight experiments. The comparison result with ground truth data from the motion capture system shows the centimeter-level precision which outperforms all the Ultra-WideBand (UWB) and even vision based method. The system is not limited by the field of view (FoV) of the camera or Global Positioning System (GPS), meanwhile on account of its estimation consistency, we believe that the proposed relative state estimation framework has the potential to be prevalently adopted by aerial swarm applications in different scenarios in multiple scales.

cs.RO↗

arXiv2020

PathVQA: 30000+ Questions for Medical Visual Question Answering

Is it possible to develop an "AI Pathologist" to pass the board-certified examination of the American Board of Pathology? To achieve this goal, the first step is to create a visual question answering (VQA) dataset where the AI agent is presented with a pathology image together with a question and is asked to give the correct answer. Our work makes the first attempt to build such a dataset. Different from creating general-domain VQA datasets where the images are widely accessible and there are many crowdsourcing workers available and capable of generating question-answer pairs, developing a medical VQA dataset is much more challenging. First, due to privacy concerns, pathology images are usually not publicly available. Second, only well-trained pathologists can understand pathology images, but they barely have time to help create datasets for AI research. To address these challenges, we resort to pathology textbooks and online digital libraries. We develop a semi-automated pipeline to extract pathology images and captions from textbooks and generate question-answer pairs from captions using natural language processing. We collect 32,799 open-ended questions from 4,998 pathology images where each question is manually checked to ensure correctness. To our best knowledge, this is the first dataset for pathology VQA. Our dataset will be released publicly to promote research in medical VQA.

cs.CL↗

arXiv2020

Review on Set-Theoretic Methods for Safety Verification and Control of Power System

Increasing penetration of renewable energy introduces significant uncertainty into power systems. Traditional simulation-based verification methods may not be applicable due to the unknown-but-bounded feature of the uncertainty sets. Emerging set-theoretic methods have been intensively investigated to tackle this challenge. The paper comprehensively reviews these methods categorized by underlying mathematical principles, that is, set operation-based methods and passivity-based methods. Set operation-based methods are more computationally efficient, while passivity-based methods provide semi-analytical expression of reachable sets, which can be readily employed for control. Other features between different methods are also discussed and illustrated by numerical examples. A benchmark example is presented and solved by different methods to verify consistency.

eess.SY↗

arXiv2020

Revealing Fine Structures of the Retinal Receptive Field by Deep Learning Networks

Deep convolutional neural networks (CNNs) have demonstrated impressive performance on many visual tasks. Recently, they became useful models for the visual system in neuroscience. However, it is still not clear what are learned by CNNs in terms of neuronal circuits. When a deep CNN with many layers is used for the visual system, it is not easy to compare the structure components of CNNs with possible neuroscience underpinnings due to highly complex circuits from the retina to higher visual cortex. Here we address this issue by focusing on single retinal ganglion cells with biophysical models and recording data from animals. By training CNNs with white noise images to predict neuronal responses, we found that fine structures of the retinal receptive field can be revealed. Specifically, convolutional filters learned are resembling biological components of the retinal circuit. This suggests that a CNN learning from one single retinal cell reveals a minimal neural network carried out in this cell. Furthermore, when CNNs learned from different cells are transferred between cells, there is a diversity of transfer learning performance, which indicates that CNNs are cell-specific. Moreover, when CNNs are transferred between different types of input images, here white noise v.s. natural images, transfer learning shows a good performance, which implies that CNNs indeed capture the full computational ability of a single retinal cell for different inputs. Taken together, these results suggest that CNNs could be used to reveal structure components of neuronal circuits, and provide a powerful model for neural system identification.

q-bio.NC↗

arXiv2020

Reconstruction of Natural Visual Scenes from Neural Spikes with Deep Neural Networks

Neural coding is one of the central questions in systems neuroscience for understanding how the brain processes stimulus from the environment, moreover, it is also a cornerstone for designing algorithms of brain-machine interface, where decoding incoming stimulus is highly demanded for better performance of physical devices. Traditionally researchers have focused on functional magnetic resonance imaging (fMRI) data as the neural signals of interest for decoding visual scenes. However, our visual perception operates in a fast time scale of millisecond in terms of an event termed neural spike. There are few studies of decoding by using spikes. Here we fulfill this aim by developing a novel decoding framework based on deep neural networks, named spike-image decoder (SID), for reconstructing natural visual scenes, including static images and dynamic videos, from experimentally recorded spikes of a population of retinal ganglion cells. The SID is an end-to-end decoder with one end as neural spikes and the other end as images, which can be trained directly such that visual scenes are reconstructed from spikes in a highly accurate fashion. Our SID also outperforms on the reconstruction of visual stimulus compared to existing fMRI decoding models. In addition, with the aid of a spike encoder, we show that SID can be generalized to arbitrary visual scenes by using the image datasets of MNIST, CIFAR10, and CIFAR100. Furthermore, with a pre-trained SID, one can decode any dynamic videos to achieve real-time encoding and decoding of visual scenes by spikes. Altogether, our results shed new light on neuromorphic computing for artificial visual systems, such as event-based visual cameras and visual neuroprostheses.

q-bio.NC↗

arXiv2020

Towards the Next Generation of Retinal Neuroprosthesis: Visual Computation with Spikes

Neuroprosthesis, as one type of precision medicine device, is aiming for manipulating neuronal signals of the brain in a closed-loop fashion, together with receiving stimulus from the environment and controlling some part of our brain/body. In terms of vision, incoming information can be processed by the brain in millisecond interval. The retina computes visual scenes and then sends its output as neuronal spikes to the cortex for further computation. Therefore, the neuronal signal of interest for retinal neuroprosthesis is spike. Closed-loop computation in neuroprosthesis includes two stages: encoding stimulus to neuronal signal, and decoding it into stimulus. Here we review some of the recent progress about visual computation models that use spikes for analyzing natural scenes, including static images and dynamic movies. We hypothesize that for a better understanding of computational principles in the retina, one needs a hypercircuit view of the retina, in which different functional network motifs revealed in the cortex neuronal network should be taken into consideration for the retina. Different building blocks of the retina, including a diversity of cell types and synaptic connections, either chemical synapses or electrical synapses (gap junctions), make the retina an ideal neuronal network to adapt the computational techniques developed in artificial intelligence for modeling of encoding/decoding visual scenes. Altogether, one needs a systems approach of visual computation with spikes to advance the next generation of retinal neuroprosthesis as an artificial visual system.

q-bio.NC↗

arXiv2019

Discovery of a Photoionized Bipolar Outflow towards the Massive Protostar G45.47+0.05

Massive protostars generate strong radiation feedback, which may help set the mass they achieve by the end of the accretion process. Studying such feedback is therefore crucial for understanding the formation of massive stars. We report the discovery of a photoionized bipolar outflow towards the massive protostar G45.47+0.05 using high-resolution observations at 1.3 mm with the Atacama Large Millimeter/Submillimeter Array (ALMA) and at 7 mm with the Karl G. Jansky Very Large Array (VLA). By modeling the free-free continuum, the ionized outflow is found to be a photoevaporation flow with an electron temperature of 10,000 K and an electron number density of ~1.5x10^7 cm^-3 at the center, launched from a disk of radius of 110 au. H30alpha hydrogen recombination line emission shows strong maser amplification, with G45 being one of very few sources to show such millimeter recombination line masers. The mass of the driving source is estimated to be 30-50 Msun based on the derived ionizing photon rate, or 30-40 Msun based on the H30alpha kinematics. The kinematics of the photoevaporated material is dominated by rotation close to the disk plane, while accelerated to outflowing motion above the disk plane. The mass loss rate of the photoevaporation outflow is estimated to be ~(2-3.5)x10^-5 Msun/yr. We also found hints of a possible jet embedded inside the wide-angle ionized outflow with non-thermal emissions. The possible co-existence of a jet and a massive photoevaporation outflow suggests that, in spite of the strong photoionization feedback, accretion is still on-going.

astro-ph.SR↗

arXiv2019

Training convolutional neural networks with cheap convolutions and online distillation

The large memory and computation consumption in convolutional neural networks (CNNs) has been one of the main barriers for deploying them on resource-limited systems. To this end, most cheap convolutions (e.g., group convolution, depth-wise convolution, and shift convolution) have recently been used for memory and computation reduction but with the specific architecture designing. Furthermore, it results in a low discriminability of the compressed networks by directly replacing the standard convolution with these cheap ones. In this paper, we propose to use knowledge distillation to improve the performance of the compact student networks with cheap convolutions. In our case, the teacher is a network with the standard convolution, while the student is a simple transformation of the teacher architecture without complicated redesigning. In particular, we propose a novel online distillation method, which online constructs the teacher network without pre-training and conducts mutual learning between the teacher and student network, to improve the performance of the student model. Extensive experiments demonstrate that the proposed approach achieves superior performance to simultaneously reduce memory and computation overhead of cutting-edge CNNs on different datasets, including CIFAR-10/100 and ImageNet ILSVRC 2012, compared to the state-of-the-art CNN compression and acceleration methods. The codes are publicly available at https://github.com/EthanZhangYC/OD-cheap-convolution.

cs.CV↗

arXiv2019

An Episodic Wide-angle Outflow in HH 46/47

During star formation, the accretion disk drives fast MHD winds which usually contain two components, a collimated jet and a radially distributed wide-angle wind. These winds entrain the surrounding ambient gas producing molecular outflows. We report recent observation of 12CO (2-1) emission of the HH 46/47 molecular outflow by the Atacama Large Millimeter/sub-millimeter Array, in which we identify multiple wide-angle outflowing shell structures in both the blue and red-shifted outflow lobes. These shells are highly coherent in position-position-velocity space, extending to >40-50 km/s in velocity and 10^4 au in space with well defined morphology and kinematics. We suggest these outflowing shells are the result of the entrainment of ambient gas by a series of outbursts from an intermittent wide-angle wind. Episodic outbursts in collimated jets are commonly observed, yet detection of a similar behavior in wide-angle winds has been elusive. Here we show clear evidence that the wide-angle component of the HH 46/47 protostellar outflows experiences similar variability seen in the collimated component.

astro-ph.GA↗

arXiv2019

High Speed Continuous Variable Source-Independent Quantum Random Number Generation

As a fundamental phenomenon in nature, randomness has a wide range of applications in the fields of science and engineering. Among different types of random number generators (RNG), quantum random number generator (QRNG) is a kind of promising RNG as it can provide provable true random numbers based on the inherent randomness of fundamental quantum processes. Nevertheless, the randomness from a QRNG can be diminished (or even destroyed) if the devices (especially the entropy source devices) are not perfect or ill-characterized. To eliminate the practical security loopholes from the source, source-independent QRNGs, which allow the source to have arbitrary and unknown dimensions, have been introduced and become one of the most important semi-device-independent QRNGs. Herein a method that enables ultra-fast unpredictable quantum random number generation from quadrature fluctuations of quantum optical field without any assumptions on the input states is proposed. Particularly, to estimate a lower bound on the extractable randomness that is independent from side information held by an eavesdropper, a new security analysis framework is established based on the extremality of Gaussian states, which can be easily extended to design and analyze new semi-device-independent continuous variable QRNG protocols. Moreover, the practical imperfections of the QRNG including the effects of excess noise, finite sampling range, finite resolution and asymmetric conjugate quadratures are taken into account and quantitatively analyzed. Finally, the proposed method is experimentally demonstrated to obtain high secure random number generation rates of 15.07 Gbits/s in off-line configuration and can potentially achieve 6 Gbits/s by real-time post-processing.

quant-ph↗

arXiv2019

Dynamics of a Massive Binary at Birth

Almost all massive stars have bound stellar companions, existing in binaries or higher-order multiples. While binarity is theorized to be an essential feature of how massive stars form, essentially all information about such properties is derived from observations of already formed stars, whose orbital properties may have evolved since birth. Little is known about binarity during formation stages. Here we report high angular resolution observations of 1.3 mm continuum and H30alpha recombination line emission, which reveal a massive protobinary with apparent separation of 180 au at the center of the massive star-forming region IRAS07299-1651. From the line-of-sight velocity difference of 9.5 km/s of the two protostars, the binary is estimated to have a minimum total mass of 18 solar masses, consistent with several other metrics, and maximum period of 570 years, assuming a circular orbit. The H30alpha line from the primary protostar shows kinematics consistent with rotation along a ring of radius of 12 au. The observations indicate that disk fragmentation at several hundred au may have formed the binary, and much smaller disks are feeding the individual protostars.

astro-ph.SR↗

arXiv2019

The SOFIA Massive (SOMA) Star Formation Survey. II. High Luminosity Protostars

We present multi-wavelength images observed with SOFIA-FORCAST from $\sim$10 to 40 $μ$m of seven high luminosity massive protostars, as part of the SOFIA Massive (SOMA) Star Formation Survey. Source morphologies at these wavelengths appear to be influenced by outflow cavities and extinction from dense gas surrounding the protostars. Using these images, we build spectral energy distributions (SEDs) of the protostars, also including archival data from Spitzer, Herschel and other facilities. Radiative transfer (RT) models of Zhang & Tan (2018), based on Turbulent Core Accretion theory, are then fit to the SEDs to estimate key properties of the protostars. Considering the best five models fit to each source, the protostars have masses $m_{*} \sim 12-64 \: M_{\odot}$ accreting at rates of $\dot{m}_{*} \sim 10^{-4}-10^{-3} \: M_{\odot} \: \rm yr^{-1}$ inside cores of initial masses $M_{c} \sim 100-500 \: M_{\odot}$ embedded in clumps with mass surface densities $Σ_{\rm cl} \sim 0.1-3 \: \rm g \: cm^{-2}$ and span a luminosity range of $10^{4} -10^{6} \: L_{\odot}$. Compared with the first eight protostars in Paper I, the sources analyzed here are more luminous, and thus likely to be more massive protostars. They are often in a clustered environment or have a companion protostar relatively nearby. From the range of parameter space of the models, we do not see any evidence that $Σ_{\rm cl}$ needs to be high to form these massive stars. For most sources the RT models provide reasonable fits to the SEDs, though the cold clump material often influences the long wavelength fitting. However, for sources in very clustered environments, the model SEDs may not be such a good description of the data, indicating potential limitations of the models for these regions.

astro-ph.GA↗

arXiv2019

An Ordered Envelope-disk Transition in the Massive Protostellar Source G339.88-1.26

We report molecular line observations of the massive protostellar source G339.88-1.26 with the Atacama Large Millimeter/Submillimeter Array. The observations reveal a highly collimated SiO jet extending from the 1.3 mm continuum source, which connects to a slightly wider but still highly collimated CO outflow. Rotational features perpendicular to the outflow axis are detected in various molecular emissions, including SiO, SO2, H2S, CH3OH, and H2CO emissions. Based on their spatial distributions and kinematics, we find that they trace different parts of the envelope-disk system. The SiO emission traces the disk and inner envelope in addition to the jet. The CH3OH and H2CO emissions mostly trace the infalling-rotating envelope, and are enhanced around the transition region between envelope and disk, i.e., the centrifugal barrier. The SO2 and H2S emissions are enhanced around the centrifugal barrier, and also trace the outer part of the disk. Envelope kinematics are consistent with rotating-infalling motion, while those of the disk are consistent with Keplerian rotation. The radius and velocity of the centrifugal barrier are estimated to be about 530 au and 6 km/s, leading to a central mass of about 11 solar masses, consistent with estimates based on spectral energy distribution fitting. These results indicate that an ordered transition from an infalling-rotating envelope to a Keplerian disk through a centrifugal barrier, accompanied by changes of types of molecular line emissions, is a valid description of this massive protostellar source. This implies that at least some massive stars form in a similar way as low-mass stars via Core Accretion.

astro-ph.GA↗