arXiv subjects

Liang Li

Publications and source records attributed to Liang Li.

At least 271 records · Page 15Linked to original sources

arXiv2022

Superconductivity in the nodal-line compound La$_3$Pt$_3$Bi$_4$

Owing to the specific topological states in nodal-line semimetals, novel topological superconductivity is expected to emerge in these systems. In this letter, by combination of the first-principles calculations and resistivity, susceptibility and specific heat measurements, we demonstrate that La$_3$Pt$_3$Bi$_4$ is a topologically nontrivial nodal-ring semimetal protected by the gliding-mirror symmetry even in the presence of spin-orbit coupling. Meanwhile, we discover bulk superconductivity with a transition temperature of $\sim$1.1 K, and an upper critical field of $\sim$0.41 T. These findings demonstrate that La$_3$Pt$_3$Bi$_4$ provides a material platform for studying novel superconductivity in the nodal-ring system.

cond-mat.supr-con↗

arXiv2022

Unsupervised Coherent Video Cartoonization with Perceptual Motion Consistency

In recent years, creative content generations like style transfer and neural photo editing have attracted more and more attention. Among these, cartoonization of real-world scenes has promising applications in entertainment and industry. Different from image translations focusing on improving the style effect of generated images, video cartoonization has additional requirements on the temporal consistency. In this paper, we propose a spatially-adaptive semantic alignment framework with perceptual motion consistency for coherent video cartoonization in an unsupervised manner. The semantic alignment module is designed to restore deformation of semantic structure caused by spatial information lost in the encoder-decoder architecture. Furthermore, we devise the spatio-temporal correlative map as a style-independent, global-aware regularization on the perceptual motion consistency. Deriving from similarity measurement of high-level features in photo and cartoon frames, it captures global semantic information beyond raw pixel-value in optical flow. Besides, the similarity measurement disentangles temporal relationships from domain-specific style properties, which helps regularize the temporal consistency without hurting style effects of cartoon images. Qualitative and quantitative experiments demonstrate our method is able to generate highly stylistic and temporal consistent cartoon videos.

cs.CV↗

arXiv2022

IR-GAN: Image Manipulation with Linguistic Instruction by Increment Reasoning

Conditional image generation is an active research topic including text2image and image translation. Recently image manipulation with linguistic instruction brings new challenges of multimodal conditional generation. However, traditional conditional image generation models mainly focus on generating high-quality and visually realistic images, and lack resolving the partial consistency between image and instruction. To address this issue, we propose an Increment Reasoning Generative Adversarial Network (IR-GAN), which aims to reason the consistency between visual increment in images and semantic increment in instructions. First, we introduce the word-level and instruction-level instruction encoders to learn user's intention from history-correlated instructions as semantic increment. Second, we embed the representation of semantic increment into that of source image for generating target image, where source image plays the role of referring auxiliary. Finally, we propose a reasoning discriminator to measure the consistency between visual increment and semantic increment, which purifies user's intention and guarantees the good logic of generated target image. Extensive experiments and visualization conducted on two datasets show the effectiveness of IR-GAN.

cs.CV↗

arXiv2022

Few Shot Generative Model Adaption via Relaxed Spatial Structural Alignment

Training a generative adversarial network (GAN) with limited data has been a challenging task. A feasible solution is to start with a GAN well-trained on a large scale source domain and adapt it to the target domain with a few samples, termed as few shot generative model adaption. However, existing methods are prone to model overfitting and collapse in extremely few shot setting (less than 10). To solve this problem, we propose a relaxed spatial structural alignment method to calibrate the target generative models during the adaption. We design a cross-domain spatial structural consistency loss comprising the self-correlation and disturbance correlation consistency loss. It helps align the spatial structural information between the synthesis image pairs of the source and target domains. To relax the cross-domain alignment, we compress the original latent space of generative models to a subspace. Image pairs generated from the subspace are pulled closer. Qualitative and quantitative experiments show that our method consistently surpasses the state-of-the-art methods in few shot setting.

cs.CV↗

arXiv2022

Open-Vocabulary One-Stage Detection with Hierarchical Visual-Language Knowledge Distillation

Open-vocabulary object detection aims to detect novel object categories beyond the training set. The advanced open-vocabulary two-stage detectors employ instance-level visual-to-visual knowledge distillation to align the visual space of the detector with the semantic space of the Pre-trained Visual-Language Model (PVLM). However, in the more efficient one-stage detector, the absence of class-agnostic object proposals hinders the knowledge distillation on unseen objects, leading to severe performance degradation. In this paper, we propose a hierarchical visual-language knowledge distillation method, i.e., HierKD, for open-vocabulary one-stage detection. Specifically, a global-level knowledge distillation is explored to transfer the knowledge of unseen categories from the PVLM to the detector. Moreover, we combine the proposed global-level knowledge distillation and the common instance-level knowledge distillation to learn the knowledge of seen and unseen categories simultaneously. Extensive experiments on MS-COCO show that our method significantly surpasses the previous best one-stage detector with 11.9\% and 6.7\% $AP_{50}$ gains under the zero-shot detection and generalized zero-shot detection settings, and reduces the $AP_{50}$ performance gap from 14\% to 7.3\% compared to the best two-stage detector.

cs.CV↗

arXiv2022

Standard GRB Spectral Models "Misused"?

The standard model characterizing the gamma-ray burst (GRB) spectrum invokes a four-parameter empirical function, the so-called the BAND model. An alternative model named cutoff power law (COMP) implements a power law with an exponential cutoff. These functions achieve almost equally good fits on observed spectra, and are adopted in nearly all of the GRB literature. Here, we reanalyze the sample defined in Li.et al.,2021,ApJS,254,35 (39 bursts including 944 spectra). We classify the spectra by two methods: (1) checking their corner-corner plots of the posteriors to determine well-constrained $β$ (BAND-better) and unconstrained $β$ (COMP-better) categories; and (2) defining the four groups by difference of the deviance information criterion (DIC). We find inconsistent peaks of the parameter distributions between the BAND-better spectra ($α=-0.64\pm0.28$ and $\rm log_{10}(E_{\rm p})=\rm log_{10}(191)\pm0.41$) and the COMP-better spectra ($α=-0.96\pm0.33$ and $\rm log_{10}(E_{\rm p})=\rm log_{10}(249)\pm0.40$). With the statistically preferred model and vice versa the misused model defined based on DIC statistics, we also find that the fitted parameters obtained by the misused model (COMP) significantly deviate from those obtained by the statistically preferred model (BAND). This means that if a spectrum is statistically preferred, described as the BAND, applying COMP to derive the spectral parameters will prominently deviate from their intrinsic shape, therefore affecting the physical interpretation. Our analysis indicates that the better or statistically preferred model should be duly examined during GRB spectral analysis. In addition, the $β$ distribution exhibits a bimodal structure containing the BAND-better and COMP-better spectra, respectively, implying that BAND and COMP both may have physical origin.

hep-ph↗

arXiv2022

Debiased Batch Normalization via Gaussian Process for Generalizable Person Re-Identification

Generalizable person re-identification aims to learn a model with only several labeled source domains that can perform well on unseen domains. Without access to the unseen domain, the feature statistics of the batch normalization (BN) layer learned from a limited number of source domains is doubtlessly biased for unseen domain. This would mislead the feature representation learning for unseen domain and deteriorate the generalizaiton ability of the model. In this paper, we propose a novel Debiased Batch Normalization via Gaussian Process approach (GDNorm) for generalizable person re-identification, which models the feature statistic estimation from BN layers as a dynamically self-refining Gaussian process to alleviate the bias to unseen domain for improving the generalization. Specifically, we establish a lightweight model with multiple set of domain-specific BN layers to capture the discriminability of individual source domain, and learn the corresponding parameters of the domain-specific BN layers. These parameters of different source domains are employed to deduce a Gaussian process. We randomly sample several paths from this Gaussian process served as the BN estimations of potential new domains outside of existing source domains, which can further optimize these learned parameters from source domains, and estimate more accurate Gaussian process by them in return, tending to real data distribution. Even without a large number of source domains, GDNorm can still provide debiased BN estimation by using the mean path of the Gaussian process, while maintaining low computational cost during testing. Extensive experiments demonstrate that our GDNorm effectively improves the generalization ability of the model on unseen domain.

cs.CV↗

arXiv2022

Modality-Adaptive Mixup and Invariant Decomposition for RGB-Infrared Person Re-Identification

RGB-infrared person re-identification is an emerging cross-modality re-identification task, which is very challenging due to significant modality discrepancy between RGB and infrared images. In this work, we propose a novel modality-adaptive mixup and invariant decomposition (MID) approach for RGB-infrared person re-identification towards learning modality-invariant and discriminative representations. MID designs a modality-adaptive mixup scheme to generate suitable mixed modality images between RGB and infrared images for mitigating the inherent modality discrepancy at the pixel-level. It formulates modality mixup procedure as Markov decision process, where an actor-critic agent learns dynamical and local linear interpolation policy between different regions of cross-modality images under a deep reinforcement learning framework. Such policy guarantees modality-invariance in a more continuous latent space and avoids manifold intrusion by the corrupted mixed modality samples. Moreover, to further counter modality discrepancy and enforce invariant visual semantics at the feature-level, MID employs modality-adaptive convolution decomposition to disassemble a regular convolution layer into modality-specific basis layers and a modality-shared coefficient layer. Extensive experimental results on two challenging benchmarks demonstrate superior performance of MID over state-of-the-art methods.

cs.CV↗

arXiv2022

Recursive Least-Squares Estimator-Aided Online Learning for Visual Tracking

Tracking visual objects from a single initial exemplar in the testing phase has been broadly cast as a one-/few-shot problem, i.e., one-shot learning for initial adaptation and few-shot learning for online adaptation. The recent few-shot online adaptation methods incorporate the prior knowledge from large amounts of annotated training data via complex meta-learning optimization in the offline phase. This helps the online deep trackers to achieve fast adaptation and reduce overfitting risk in tracking. In this paper, we propose a simple yet effective recursive least-squares estimator-aided online learning approach for few-shot online adaptation without requiring offline training. It allows an in-built memory retention mechanism for the model to remember the knowledge about the object seen before, and thus the seen data can be safely removed from training. This also bears certain similarities to the emerging continual learning field in preventing catastrophic forgetting. This mechanism enables us to unveil the power of modern online deep trackers without incurring too much extra computational cost. We evaluate our approach based on two networks in the online learning families for tracking, i.e., multi-layer perceptrons in RT-MDNet and convolutional neural networks in DiMP. The consistent improvements on several challenging tracking benchmarks demonstrate its effectiveness and efficiency.

cs.CV↗

arXiv2022

Fingerprint of the Interbond Electron Hopping in Second-Order Harmonic Generation

We experimentally explore the fingerprint of the microscopic electron dynamics in second-order harmonic generation (SHG). It is shown that the interbond electron hopping induces a novel source of nonlinear polarization and plays an important role even when the driving laser intensity is 2 orders of magnitude lower than the characteristic atomic field. Our model predicts anomalous anisotropic structures of the SHG yield contributed by the interbond electron hopping, which is identified in our experiments with ZnO crystals. Moreover, a generalized second-order susceptibility with an explicit form is proposed, which provides a unified description in both the weak and strong field regimes. Our work reveals the nonlinear responses of materials at the electron scale and extends the nonlinear optics to a previously unexplored regime, where the nonlinearity related to the interbond electron hopping becomes dominant. It paves the way for realizing controllable nonlinearity on an ultrafast time scale.

physics.optics↗

arXiv2022

The concentration of zero-noise limits of invariant measures for stochastic dynamical systems

In this paper, we study concentration phenomena of zero-noise limits of invariant measures for stochastic differential equations defined on $\mathbb{R}^d$ with locally Lipschitz continuous coefficients and more than one ergodic state. Under some dissipative conditions, by using Lyapunov-like functions and large deviations methods, we estimate the invariant measures in neighborhoods of stable sets, neighborhoods of unstable sets and their complement, respectively. Our result illustrates that invariant measures concentrate on the intersection of stable sets where a cost functional $W(K_i)$ is minimized and the Birkhoff center of the corresponding deterministic systems as noise tends down to zero. Furthermore, we prove the large deviations principle of invariant measures. At the end of this paper, we provide some explicit examples and their numerical simulations.

math.PR↗

arXiv2022

The white dwarf binary merger model of GRB 170817A

Following the GRB 170817A prompt emission lasting a fraction of a second, $10^8$ s of data in the X-rays, optical, and radio wavelengths have been acquired. We here present a model that fits the spectra, flux, and time variability of all these emissions, based on the thermal and synchrotron cooling of the expanding matter ejected in a binary white dwarf merger. The $10^{-3} M_\odot$ of ejecta, expanding at velocities of $10^9$ cm s$^{-1}$, are powered by the newborn massive, fast rotating, magnetized white dwarf with a mass of $1.3 M_\odot$, a rotation period of $\gtrsim 12$ s, and a dipole magnetic field $\sim 10^{10}$ G, born in the merger of a $1.0+0.8 M_\odot$ white dwarf binary. Therefore, the long-lasting mystery of the GRB 170817A nature is solved by the merger of a white dwarf binary that also explains the prompt emission energetics.

astro-ph.HE↗

arXiv2022

Quantum photonics in triangular-cross-section nanodevices in silicon carbide

Silicon carbide is evolving as a prominent solid-state platform for the realization of quantum information processing hardware. Angle-etched nanodevices are emerging as a solution to photonic integration in bulk substrates where color centers are best defined. We model triangular cross-section waveguides and photonic crystal cavities using Finite-Difference Time-Domain and Finite-Difference Eigensolver approaches. We analyze optimal color center positioning within the modes of these devices and provide estimates on achievable Purcell enhancement in nanocavities with applications in quantum communications. Using open quantum system modeling, we explore emitter-cavity interactions of multiple non-identical color centers coupled to both a single cavity and a photonic crystal molecule in SiC. We observe polariton and subradiant state formation in the cavity-protected regime of cavity quantum electrodynamics applicable in quantum simulation.

physics.optics↗

arXiv2022

Is magnetically dominated outflow required to explain GRBs?

The composition of relativistic outflows producing gamma-ray bursts is a long standing open question. One of the main arguments in favor of magnetically dominated outflows is the absence of photospheric component in their broadband time resolved spectra, with such notable example as GRB 080916C. Here, we perform a time-resolved analysis of this burst and confirm the previous detection of an additional spectral component. We show that this subdominant component is consistent with the photosphere of ultrarelativistic baryonic outflow, deep in the coasting regime. We argue that, contrary to previous statements, the magnetic dominance of the outflow is not required for the interpretation of this GRB. Moreover, simultaneous detection of high energy emission in its prompt phase requires departure from a one-zone emission model.

astro-ph.HE↗

arXiv2022

Asymptotic behaviour of non-radiative solution to the wave equations

In this work we consider weakly non-radiative solutions to both linear and non-linear wave equations. We first characterize all weakly non-radiative free waves, without the radial assumption. Then in dimension 3 we show that the initial data of non-radiative solutions to a wide range of nonlinear wave equations are similar to those of non-radiative free waves in term of asymptotic behaviour.

math.AP↗

arXiv2021

Complex contraction on trees without proof of correlation decay

We prove complex contraction for zero-free regions of counting weighted set cover problem in which an element can appear in an unbounded number of sets, thus obtaining fully polynomial-time approximation schemes(FPTAS) via Barvinok's algorithmic paradigm\cite{barvinok2016combinatorics}. Relying on the computation tree expansion, our approach does not need proof of correlation decay in the real axis. We directly look in the complex plane for a region that contracts into its interior as the tree recursion procedure goes from leaves to the root. For the class of problems under the framework of weighted set covers, we are able to give a general approach for describing the contraction regions and draw a unified algorithmic conclusion. Several previous results, including counting (weighted-)edge covers, counting bipartite independent sets and counting monotone CNFs can be completely or partially covered by our main theorem. In contrast to the correlation decay method which also depends on tree expansions and needs different potential functions for different problems, our approach is more generic in the sense that our contraction region for different problems shares a common shape in the complex plane.

cs.DS↗

arXiv2021

Testing High-latitude Curvature Effect of Gamma-Ray Bursts with {\it Fermi} Data: Evidence of Bulk Acceleration in Prompt Emission

When a gamma-ray burst (GRB) emitter stops emission abruptly, the observer receives rapidly fading emission from high latitudes with respect to the line of sight, known as the ``curvature effect''. Identifying such emission from GRB prompt-emission lightcurves would constrain the radius of prompt emission from the central engine and the composition of GRB jets. We perform a dedicated search of high-latitude emission (HLE) through spectral and temporal analyses of a sample of single-pulse bursts detected by the Gamma-ray Burst Monitor on board the {\it Fermi} satellite. We identify HLE from a subsample of bursts and constrain the emission radius to be $R_{\rm GRB} \sim (10^{15}-10^{16})$ cm from the central engine. Some bursts have the HLE decay faster than predicted by a constant Lorentz factor jet, suggesting that the emission region is undergoing acceleration during prompt emission. This supports the Poynting-flux-dominated jet composition for these bursts. The conclusion is consistent with previous results drawn from spectral-lag modeling of prompt emission and HLE analysis of X-ray flares.

astro-ph.HE↗

arXiv2021

Bayesian Time-Resolved Spectroscopy of Multi-Pulsed GRBs: Variations of Emission Properties amongst Pulses

Gamma-ray bursts (GRBs) are highly variable and exhibit strong spectral evolution. In particular, the emission properties vary from pulse to pulse in multipulse bursts. Here we present a time-resolved Bayesian spectral analysis of a compilation of GRB pulses observed by the {\it Fermi}/Gamma-ray Burst Monitor. The pulses are selected to have at least four timebins with a high statistical significance, which ensures that the spectral fits are well determined and that spectral correlations can be established. The sample consists of 39 bursts, 117 pulses, and 1228 spectra. We confirm the general trend that pulses become softer over time, with mainly the low-energy power-law index $α$ becoming smaller. A few exceptions to this trend exist, with the hardest pulse occurring at late times. The first pulse in a burst is clearly different from the later pulses; three-fourths of them violate the synchrotron line of death, while around half of them significantly prefer photospheric emission. These fractions decrease for subsequent pulses. We also find that in two-thirds of the pulses, the spectral parameters ($α$ and peak energy) track the light-curve variations. This is a larger fraction compared to what is found in previous samples. In conclusion, emission compatible with the GRB photosphere is typically found close to the trigger time, while the chance of detecting synchrotron emission is greatest at late times. This allows for the coexistence of emission mechanisms at late times.

astro-ph.HE↗