arXiv subjects

Jian Chen

Publications and source records attributed to Jian Chen.

At least 307 records · Page 17Linked to original sources

arXiv2020

Hierarchical Neural Architecture Search for Single Image Super-Resolution

Deep neural networks have exhibited promising performance in image super-resolution (SR). Most SR models follow a hierarchical architecture that contains both the cell-level design of computational blocks and the network-level design of the positions of upsampling blocks. However, designing SR models heavily relies on human expertise and is very labor-intensive. More critically, these SR models often contain a huge number of parameters and may not meet the requirements of computation resources in real-world applications. To address the above issues, we propose a Hierarchical Neural Architecture Search (HNAS) method to automatically design promising architectures with different requirements of computation cost. To this end, we design a hierarchical SR search space and propose a hierarchical controller for architecture search. Such a hierarchical controller is able to simultaneously find promising cell-level blocks and network-level positions of upsampling layers. Moreover, to design compact architectures with promising performance, we build a joint reward by considering both the performance and computation cost to guide the search process. Extensive experiments on five benchmark datasets demonstrate the superiority of our method over existing methods.

cs.CV↗

arXiv2020

Closed-loop Matters: Dual Regression Networks for Single Image Super-Resolution

Deep neural networks have exhibited promising performance in image super-resolution (SR) by learning a nonlinear mapping function from low-resolution (LR) images to high-resolution (HR) images. However, there are two underlying limitations to existing SR methods. First, learning the mapping function from LR to HR images is typically an ill-posed problem, because there exist infinite HR images that can be downsampled to the same LR image. As a result, the space of the possible functions can be extremely large, which makes it hard to find a good solution. Second, the paired LR-HR data may be unavailable in real-world applications and the underlying degradation method is often unknown. For such a more general case, existing SR models often incur the adaptation problem and yield poor performance. To address the above issues, we propose a dual regression scheme by introducing an additional constraint on LR data to reduce the space of the possible functions. Specifically, besides the mapping from LR to HR images, we learn an additional dual regression mapping estimates the down-sampling kernel and reconstruct LR images, which forms a closed-loop to provide additional supervision. More critically, since the dual regression process does not depend on HR images, we can directly learn from LR images. In this sense, we can easily adapt SR models to real-world data, e.g., raw video frames from YouTube. Extensive experiments with paired training data and unpaired real-world data demonstrate our superiority over existing methods.

cs.CV↗

arXiv2020

Neural encoding and interpretation for high-level visual cortices based on fMRI using image caption features

On basis of functional magnetic resonance imaging (fMRI), researchers are devoted to designing visual encoding models to predict the neuron activity of human in response to presented image stimuli and analyze inner mechanism of human visual cortices. Deep network structure composed of hierarchical processing layers forms deep network models by learning features of data on specific task through big dataset. Deep network models have powerful and hierarchical representation of data, and have brought about breakthroughs for visual encoding, while revealing hierarchical structural similarity with the manner of information processing in human visual cortices. However, previous studies almost used image features of those deep network models pre-trained on classification task to construct visual encoding models. Except for deep network structure, the task or corresponding big dataset is also important for deep network models, but neglected by previous studies. Because image classification is a relatively fundamental task, it is difficult to guide deep network models to master high-level semantic representations of data, which causes into that encoding performance for high-level visual cortices is limited. In this study, we introduced one higher-level vision task: image caption (IC) task and proposed the visual encoding model based on IC features (ICFVEM) to encode voxels of high-level visual cortices. Experiment demonstrated that ICFVEM obtained better encoding performance than previous deep network models pre-trained on classification task. In addition, the interpretation of voxels was realized to explore the detailed characteristics of voxels based on the visualization of semantic words, and comparative analysis implied that high-level visual cortices behaved the correlative representation of image content.

cs.CV↗

arXiv2020

A Gd@C82-based single molecular electret device with switchable electrical polarization

Single molecular electrets exhibiting single molecule electric polarization switching have been long desired as a platform for extremely small non-volatile storage devices, although it is controversial because of the poor stability of single molecular electric dipoles. Here we study the single molecular device of GdC82, where the encapsulated Gd atom forms a charge center, and we have observed a gate controlled switching behavior between two sets of single electron transport stability diagrams. The switching is operated in a hysteresis loop with a coercive gate field of around 0.5Vnm. Theoretical calculations have assigned the two conductance diagrams to corresponding energy levels of two states that the Gd atom is trapped at two different sites of the C82 cage, which possess two different permanent electrical dipole orientations. The two dipole states are stabilized by the anisotropic energy and separated by a transition energy barrier of 70 meV. Such switching is then accessed to the electric field driven reorientation of individual dipole while overcoming the barriers by the coercive gate field, and demonstrates the creation of a single molecular electret.

cond-mat.mes-hall↗

arXiv2020

Core-level x-ray photoemission and Raman spectroscopy studies on electronic structures in Mott-Hubbard type nickelate oxide NdNiO$_2$

We perform core-level X-ray photoemission spectroscopy (XPS) and electronic Raman scattering studies of electronic structures and spin fluctuations in the bulk samples of the nickelate oxide NdNiO$_2$. According to Nd $3d$ and O $1s$ XPS spectra, we conclude that NdNiO$_2$ has a large transfer energy. From the analysis of the main line of the Ni $2p_{3/2}$ XPS, we confirm the NiO$_2$ planes in NdNiO$_2$ are of Mott-Hubbard type in the Zaanen-Sawatzky-Allen scheme. The two-magnon peak in the Raman scattering provides direct evidence for the strong spin-fluctuation in NdNiO$_2$. The peak position determines the antiferromagnetic exchange $J=25$~meV. Our experimental results agree well with our previous theoretical results.

cond-mat.supr-con↗

arXiv2020

BigGAN-based Bayesian reconstruction of natural images from human brain activity

In the visual decoding domain, visually reconstructing presented images given the corresponding human brain activity monitored by functional magnetic resonance imaging (fMRI) is difficult, especially when reconstructing viewed natural images. Visual reconstruction is a conditional image generation on fMRI data and thus generative adversarial network (GAN) for natural image generation is recently introduced for this task. Although GAN-based methods have greatly improved, the fidelity and naturalness of reconstruction are still unsatisfactory due to the small number of fMRI data samples and the instability of GAN training. In this study, we proposed a new GAN-based Bayesian visual reconstruction method (GAN-BVRM) that includes a classifier to decode categories from fMRI data, a pre-trained conditional generator to generate natural images of specified categories, and a set of encoding models and evaluator to evaluate generated images. GAN-BVRM employs the pre-trained generator of the prevailing BigGAN to generate masses of natural images, and selects the images that best matches with the corresponding brain activity through the encoding models as the reconstruction of the image stimuli. In this process, the semantic and detailed contents of reconstruction are controlled by decoded categories and encoding models, respectively. GAN-BVRM used the Bayesian manner to avoid contradiction between naturalness and fidelity from current GAN-based methods and thus can improve the advantages of GAN. Experimental results revealed that GAN-BVRM improves the fidelity and naturalness, that is, the reconstruction is natural and similar to the presented image stimuli.

cs.CV↗

arXiv2020

One-Shot Parameter Identification of the Thevenin's Model for Batteries: Methods and Validation

Parameter estimation is of foundational importance for various model-based battery management tasks, including charging control, state-of-charge estimation and aging assessment. However, it remains a challenging issue as the existing methods generally depend on cumbersome and time-consuming procedures to extract battery parameters from data. Departing from the literature, this paper sets the unique aim of identifying all the parameters offline in a one-shot procedure, including the resistance and capacitance parameters and the parameters in the parameterized function mapping from the state-of-charge to the open-circuit voltage. Considering the well-known Thevenin's battery model, the study begins with the parameter identifiability analysis, showing that all the parameters are locally identifiable. Then, it formulates the parameter identification problem in a prediction-error-minimization framework. As the non-convexity intrinsic to the problem may lead to physically meaningless estimates, two methods are developed to overcome this issue. The first one is to constrain the parameter search within a reasonable space by setting parameter bounds, and the other adopts regularization of the cost function using prior parameter guess. The proposed identifiability analysis and identification methods are extensively validated through simulations and experiments.

eess.SY↗

arXiv2020

Nonlinear Double-Capacitor Model for Rechargeable Batteries: Modeling, Identification and Validation

This paper proposes a new equivalent circuit model for rechargeable batteries by modifying a double-capacitor model proposed in [1]. It is known that the original model can address the rate capacity effect and energy recovery effect inherent to batteries better than other models. However, it is a purely linear model and includes no representation of a battery's nonlinear phenomena. Hence, this work transforms the original model by introducing a nonlinear-mapping-based voltage source and a serial RC circuit. The modification is justified by an analogy with the single-particle model. Two parameter estimation approaches, termed 1.0 and 2.0, are designed for the new model to deal with the scenarios of constant-current and variable-current charging/discharging, respectively. In particular, the 2.0 approach proposes the notion of Wiener system identification based on maximum a posteriori estimation, which allows all the parameters to be estimated in one shot while overcoming the nonconvexity or local minima issue to obtain physically reasonable estimates. An extensive experimental evaluation shows that the proposed model offers excellent accuracy and predictive capability. A comparison against the Rint and Thevenin models further points to its superiority. With high fidelity and low mathematical complexity, this model is beneficial for various real-time battery management applications.

eess.SY↗

arXiv2020

The Shallow End: Empowering Shallower Deep-Convolutional Networks through Auxiliary Outputs

Depth is one of the key factors behind the success of convolutional neural networks (CNNs). Since ResNet, we are able to train very deep CNNs as the gradient vanishing issue has been largely addressed by the introduction of skip connections. However, we observe that, when the depth is very large, the intermediate layers (especially shallow layers) may fail to receive sufficient supervision from the loss due to the severe transformation through a long backpropagation path. As a result, the representation power of intermediate layers can be very weak and the model becomes very redundant with limited performance. In this paper, we first investigate the supervision vanishing issue in existing backpropagation (BP) methods. And then, we propose to address it via an effective method, called Multi-way BP (MW-BP), which relies on multiple auxiliary losses added to the intermediate layers of the network. The proposed MW-BP method can be applied to most deep architectures with slight modifications, such as ResNet and MobileNet. Our method often gives rise to much more compact models (denoted by "Mw+Architecture") than existing methods. For example, MwResNet-44 with 44 layers performs better than ResNet-110 with 110 layers on CIFAR-10 and CIFAR-100. More critically, the resultant models even outperform the light models obtained by state-of-the-art model compression methods. Last, our method inherently produces multiple compact models with different depths at the same time, which is helpful for model selection.

cs.CV↗

arXiv2020

Deep Learning for Image Super-resolution: A Survey

Image Super-Resolution (SR) is an important class of image processing techniques to enhance the resolution of images and videos in computer vision. Recent years have witnessed remarkable progress of image super-resolution using deep learning techniques. This article aims to provide a comprehensive survey on recent advances of image super-resolution using deep learning approaches. In general, we can roughly group the existing studies of SR techniques into three major categories: supervised SR, unsupervised SR, and domain-specific SR. In addition, we also cover some other important issues, such as publicly available benchmark datasets and performance evaluation metrics. Finally, we conclude this survey by highlighting several future directions and open issues which should be further addressed by the community in the future.

cs.CV↗

arXiv2020

NAT: Neural Architecture Transformer for Accurate and Compact Architectures

Designing effective architectures is one of the key factors behind the success of deep neural networks. Existing deep architectures are either manually designed or automatically searched by some Neural Architecture Search (NAS) methods. However, even a well-searched architecture may still contain many non-significant or redundant modules or operations (e.g., convolution or pooling), which may not only incur substantial memory consumption and computation cost but also deteriorate the performance. Thus, it is necessary to optimize the operations inside an architecture to improve the performance without introducing extra computation cost. Unfortunately, such a constrained optimization problem is NP-hard. To make the problem feasible, we cast the optimization problem into a Markov decision process (MDP) and seek to learn a Neural Architecture Transformer (NAT) to replace the redundant operations with the more computationally efficient ones (e.g., skip connection or directly removing the connection). Based on MDP, we learn NAT by exploiting reinforcement learning to obtain the optimization policies w.r.t. different architectures. To verify the effectiveness of the proposed strategies, we apply NAT on both hand-crafted architectures and NAS based architectures. Extensive experiments on two benchmark datasets, i.e., CIFAR-10 and ImageNet, demonstrate that the transformed architecture by NAT significantly outperforms both its original form and those architectures optimized by existing methods.

cs.LG↗

arXiv2019

Human Gist Processing Augments Deep Learning Breast Cancer Risk Assessment

Radiologists can classify a mammogram as normal or abnormal at better than chance levels after less than a second's exposure to the images. In this work, we combine these radiologists' gist inputs into pre-trained machine learning models to validate that integrating gist with a CNN model can achieve an AUC (area under the curve) statistically significantly higher than either the gist perception of radiologists or the model without gist input.

physics.med-ph↗

arXiv2019

Photonic Cyclone: spatiotemporal optical vortex with controllable transverse orbital angular momentum

Today, it is well known that light possesses a linear momentum which is along the propagation direction. Besides, scientists also discovered that light can possess an angular momentum (AM), a spin angular momentum (SAM) associated with circular polarization and an orbital angular momentum (OAM) owing to the azimuthally dependent phase. Even though such angular momenta are longitudinal in general, a SAM transverse to the propagation has opened up a variety of key applications [1]. In contrast, investigations of the transverse OAM are quite rare due to its complex nature. Here we demonstrate a simple method to generate a three dimensional (3D) optical wave packet with a controllable purely transverse OAM. Such a wave packet is a spatiotemporal (ST) vortex, which resembles an advancing cyclone, with optical energy flowing in the spatial and temporal dimension. Contrary to the transverse SAM, the magnitude of the transverse OAM carried by the photonic cyclone is scalable to a larger value by simple adjustments. Since the ST vortex carries a controllable OAM in the unique transverse dimension, it has a strong potential for novel applications that may not be possible otherwise. The scheme reported here can be readily adapted for the other spectra regime and different wave fields, opening tremendous opportunities for the study and applications of ST vortex in much broader scopes.

physics.optics↗

arXiv2019

Interlayer Topological Transport and Devices based on Layer Pseudospins in Photonic Valley-Hall Phases

Valley-Hall phases, first proposed in two-dimensional (2D) materials, originate from nontrivial topologies around valleys which denote local extrema in momentum space. Since they are extended into classical systems, their designs draw inspirations from existing quantum counterparts, and their transports show similar topological protections. In contrast, it is recently established in acoustics that layer pseudospins in valley-Hall phases can give rise to special valley-Hall edge states with fundamentally different transport behaviors at the interfaces compared with various 2D materials. Their realization in other classical systems, such as photonics, would allow us to design topological insulators beyond quantum inspirations. In this work, we show that layer pseudospins exist in photonic valley-Hall phases, using vertically coupled designer surface plasmon crystals, a non-radiative system in open environment supporting tightly-confined propagating modes. The negligible thermal and radiative losses in our structure pave the way for our direct observations of the layer pseudospins and associated topological phenomena stemmed from them in both real and reciprocal spaces. Photonic devices that manipulate signals based on layer pseudospins of the topological phases, such as layer convertors and layer-selected delay lines, are experimentally demonstrated, confirming the potential applications of layer pseudospins as a new degree of freedom carrying information.

physics.optics↗

arXiv2019

Effective and efficient ROI-wise visual encoding using an end-to-end CNN regression model and selective optimization

Recently, visual encoding based on functional magnetic resonance imaging (fMRI) have realized many achievements with the rapid development of deep network computation. Visual encoding model is aimed at predicting brain activity in response to presented image stimuli. Currently, visual encoding is accomplished mainly by firstly extracting image features through convolutional neural network (CNN) model pre-trained on computer vision task, and secondly training a linear regression model to map specific layer of CNN features to each voxel, namely voxel-wise encoding. However, the two-step manner model, essentially, is hard to determine which kind of well features are well linearly matched for beforehand unknown fMRI data with little understanding of human visual representation. Analogizing computer vision mostly related human vision, we proposed the end-to-end convolution regression model (ETECRM) in the region of interest (ROI)-wise manner to accomplish effective and efficient visual encoding. The end-to-end manner was introduced to make the model automatically learn better matching features to improve encoding performance. The ROI-wise manner was used to improve the encoding efficiency for many voxels. In addition, we designed the selective optimization including self-adapting weight learning and weighted correlation loss, noise regularization to avoid interfering of ineffective voxels in ROI-wise encoding. Experiment demonstrated that the proposed model obtained better predicting accuracy than the two-step manner of encoding models. Comparative analysis implied that end-to-end manner and large volume of fMRI data may drive the future development of visual encoding.

q-bio.NC↗

arXiv2019

What Do People See in a Twenty-Second Glimpse of Bivariate Vector Field Visualizations?

Little is known about how people learn from a brief glimpse of three-dimensional (3D) bivariate vector field visualizations and about how well visual features can guide behavior. Here we report empirical study results on the use of color, texture, and length to guide viewing of bivariate glyphs: these three visual features are mapped to the first integer variable (v1) and length to the second quantitative variable (v2). Participants performed two tasks within 20 seconds: (1) MAX: find the largest v2 when v1 is fixed; (2) SEARCH: find a specific bivariate variable shown on the screen in a vector field. Our first study with eighteen participants performing these tasks showed that the randomized vector positions, although they lessened viewers' ability to group vectors, did not reduce task accuracy compared to structured vector fields. This result may support that these color, texture, and length can provide to a certain degree, guide viewers' attention to task-relevant regions. The second study measured eye movement to quantify viewers' behaviors with three-errors (scanning, recognition, and decision errors) and one-behavior (refixation) metrics. Our results showed two dominant search strategies: drilling and scanning. Coloring tended to restrict eye movement to the task-relevant regions of interest, enabling drilling. Length tended to support scanners who quickly wandered around at different v1 levels. Drillers had significantly less errors than scanners and the error rates for color and texture were also lowest. And length had limited discrimination power than color and texture as a 3D visual guidance. Our experiment results may suggest that using categorical visual feature could help obtain the global structure of a vector field visualization. We provide the first benchmark of the attention cost of seeing a bivariate vector on average about 5 items per second.

cs.GR↗

arXiv2019

Picturing Bivariate Separable-Features for Univariate Vector Magnitudes in Large-Magnitude-Range Quantum Physics Data

We present study results from two experiments to empirically validate that separable bivariate pairs for univariate representations of large-magnitude-range vectors are more efficient than integral pairs. The first experiment with 20 participants compared: one integral pair, three separable pairs, and one redundant pair, which is a mix of the integral and separable features. Participants performed three local tasks requiring reading numerical values, estimating ratio, and comparing two points. The second 18-participant study compared three separable pairs using three global tasks when participants must look at the entire field to get an answer: find a specific target in 20 seconds, find the maximum magnitude in 20 seconds, and estimate the total number of vector exponents within 2 seconds. Our results also reveal the following: separable pairs led to the most accurate answers and the shortest task execution time, while integral dimensions were among the least accurate; it achieved high performance only when a pop-out separable feature (here color) was added. To reconcile this finding with the existing literature, our second experiment suggests that the higher the separability, the higher the accuracy; the reason is probably that the emergent global scene created by the separable pairs reduces the subsequent search space.

cs.GR↗

arXiv2019

Cycle-Consistent Adversarial GAN: the integration of adversarial attack and defense

In image classification of deep learning, adversarial examples where inputs intended to add small magnitude perturbations may mislead deep neural networks (DNNs) to incorrect results, which means DNNs are vulnerable to them. Different attack and defense strategies have been proposed to better research the mechanism of deep learning. However, those research in these networks are only for one aspect, either an attack or a defense, not considering that attacks and defenses should be interdependent and mutually reinforcing, just like the relationship between spears and shields. In this paper, we propose Cycle-Consistent Adversarial GAN (CycleAdvGAN) to generate adversarial examples, which can learn and approximate the distribution of original instances and adversarial examples. For CycleAdvGAN, once the Generator and are trained, can generate adversarial perturbations efficiently for any instance, so as to make DNNs predict wrong, and recovery adversarial examples to clean instances, so as to make DNNs predict correct. We apply CycleAdvGAN under semi-white box and black-box settings on two public datasets MNIST and CIFAR10. Using the extensive experiments, we show that our method has achieved the state-of-the-art adversarial attack method and also efficiently improve the defense ability, which make the integration of adversarial attack and defense come true. In additional, it has improved attack effect only trained on the adversarial dataset generated by any kind of adversarial attack.

cs.CV↗