arXiv subjects

Liang Li

Publications and source records attributed to Liang Li.

At least 235 records · Page 13Linked to original sources

arXiv2023

Revisiting the Spectral-Energy Correlations of GRBs with {\it Fermi} Data I: Model-wise Properties

Gamma-ray bursts (GRBs) exhibit a diversity of spectra. Several spectral models (e.g., Band, cutoff power-law, and blackbody) and their hybrid versions (e.g., Band+blackbody) have been widely used to fit the observed GRB spectra. Here, we attempt to collect all the bursts detected by {\it Fermi}-GBM with known redshifts from July 2008 to May 2022, motivated to (i) provide a parameter catalog independent from the official \emph{Fermi}/GBM team and (ii) achieve a ``clean" model-based GRB spectral-energy correlation analysis. A nearly complete GRB sample was created, containing 153 such bursts (136 long gamma-ray bursts and 17 short gamma-ray bursts). Using the sample and by performing detailed spectral analysis and model comparisons, we investigate two GRB spectral-energy correlations: the cosmological rest-frame peak energy ($E_{\rm p,z}$) of the $νF_ν$ prompt emission spectrum correlated with (i) the isotropic-bolometric-equivalent emission energy $E_{γ, \rm iso}$ (the Amati relation), and (ii) the isotropic-bolometric-equivalent peak luminosity $L_{\rm p, iso}$ (the Yonetoku relation). From a linear regression analysis, a tight correlation between $E_{\rm p,z}$ and $E_{γ, \rm iso}$ (and $L_{γ,\rm iso}$) is found for both the Band-like and CPL-like bursts. More interestingly, the CPL-like bursts do not fall on the Band-like burst Amati and Yonetoku correlations, suggesting distinct radiation processes, and pointing towards the fact that these spectral-energy correlations are tightly reliant on the model-wise properties.

astro-ph.HE↗

arXiv2023

Neighborhood Contrastive Transformer for Change Captioning

Change captioning is to describe the semantic change between a pair of similar images in natural language. It is more challenging than general image captioning, because it requires capturing fine-grained change information while being immune to irrelevant viewpoint changes, and solving syntax ambiguity in change descriptions. In this paper, we propose a neighborhood contrastive transformer to improve the model's perceiving ability for various changes under different scenes and cognition ability for complex syntax structure. Concretely, we first design a neighboring feature aggregating to integrate neighboring context into each feature, which helps quickly locate the inconspicuous changes under the guidance of conspicuous referents. Then, we devise a common feature distilling to compare two images at neighborhood level and extract common properties from each image, so as to learn effective contrastive information between them. Finally, we introduce the explicit dependencies between words to calibrate the transformer decoder, which helps better understand complex syntax structure during training. Extensive experimental results demonstrate that the proposed method achieves the state-of-the-art performance on three public datasets with different change scenarios. The code is available at https://github.com/tuyunbin/NCT.

cs.CV↗

arXiv2023

Age of Information of CSMA/CA Based Wireless Networks

We consider a wireless network where N nodes compete for a shared channel over the CSMA/CA protocol to deliver observed updates to a common remote monitor. For this network, we rate the information freshness of the CSMA/CA based network using the age of information (AoI). Different from previous work, the network we consider is unsaturated. To theoretically analyze the transmission behavior of the CSMA/CA based network, we, therefore, develop an equivalent and tractable Markov transmission model. Based on this newly developed model, the transmission probability, collision probability and average AoI of the network are obtained. Our numerical results show that as the packet rate and the number of nodes increase, both the transmission probability and collision probability are increasing; the average AoI first decreases and then increases as the packet rate increases and increases with the number of nodes.

cs.IT↗

arXiv2023

Plan-then-Seam: Towards Efficient Table-to-Text Generation

Table-to-text generation aims at automatically generating text to help people conveniently obtain salient information in tables. Recent works explicitly decompose the generation process into content planning and surface generation stages, employing two autoregressive networks for them respectively. However, they are computationally expensive due to the non-parallelizable nature of autoregressive decoding and the redundant parameters of two networks. In this paper, we propose the first totally non-autoregressive table-to-text model (Plan-then-Seam, PTS) that produces its outputs in parallel with one single network. PTS firstly writes and calibrates one plan of the content to be generated with a novel rethinking pointer predictor, and then takes the plan as the context for seaming to decode the description. These two steps share parameters and perform iteratively to capture token inter-dependency while keeping parallel decoding. Experiments on two public benchmarks show that PTS achieves 3.0~5.6 times speedup for inference time, reducing 50% parameters, while maintaining as least comparable performance against strong two-stage table-to-text competitors.

cs.CL↗

arXiv2023

A Cosmological Fireball with Sixteen-Percent Gamma-Ray Radiative Efficiency

Gamma-ray bursts (GRBs) are the most powerful explosions in the universe. How efficiently the jet converts its energy to radiation is a long-standing problem and it is poorly constrained. The standard model invokes a relativistic fireball with a bright photosphere emission component. A definitive diagnosis of GRB radiation components and measurement of GRB radiative efficiency require prompt emission and afterglow data with high-resolution and wide-band coverage in time and energy. Here we report a comprehensive temporal and spectral analysis of the TeV-emitting bright GRB 190114C. Its fluence is one of the highest of all GRBs detected so far, which allows us to perform a high-resolution study of the prompt emission spectral properties and their temporal evolution down to a timescale of about 0.1 s. We observe that each of the initial pulses has a thermal component contributing $\sim20\%$ of the total energy, the corresponding temperature and the inferred Lorentz factor of the photosphere evolve following broken power-law shapes. From the observation of the non-thermal spectra and the light curve, the onset of afterglow corresponding to the deceleration of the fireball is considered at $\sim 6$~s. By incorporating the thermal and the non-thermal observations, as well as the photosphere and the synchrotron radiative mechanisms, we can directly derive the fireball energy budget with little dependence on hypothetical parameters and to measure a $\sim 16\%$ radiative efficiency for this GRB. With the fireball energy budget derived, the afterglow microphysics parameters can also be constrained directly from the data.

astro-ph.HE↗

arXiv2023

A simple cognitive model explains movement decisions during schooling in zebrafish

While moving, animals must frequently make decisions about their future travel direction, whether they are alone or in a group. Here we investigate this process for zebrafish (Danio rerio), which naturally move in cohesive groups. Employing state-of-the-art virtual reality, we study how real fish follow one or several moving, virtual conspecifics. These data are used to inform, and test, a model of social response that includes a process of explicit decision-making, whereby the fish can decide which of the virtual conspecifics to follow, or to follow some average direction. This approach is in contrast with previous models where the direction of motion was based on a continuous computation, such as directional averaging. Building upon a simplified version of this model [Sridhar et al. 2021], which was limited to a one-dimensional projection of the fish motion, we present here a model that describes the motion of the real fish as it swims freely in two-dimensions. Motivated by experimental observations, the swim speed of the fish in this model uses a burst-and-coast swimming pattern, with the burst frequency being dependent on the distance of the fish from the followed conspecific(s). We demonstrate that this model is able to explain the observed spatial distribution of the real fish behind the virtual conspecifics in the experiments. In particular, the model naturally explains the observed critical bifurcations for a freely swimming fish, whenever the fish makes a decision to follow only one of the virtual conspecifics, instead of following them as an averaged group. This model can provide the foundation for modeling a cohesive shoal of swimming fish, while explicitly describing their directional decision-making process at the individual level

physics.bio-ph↗

arXiv2023

Hard Sample Aware Network for Contrastive Deep Graph Clustering

Contrastive deep graph clustering, which aims to divide nodes into disjoint groups via contrastive mechanisms, is a challenging research spot. Among the recent works, hard sample mining-based algorithms have achieved great attention for their promising performance. However, we find that the existing hard sample mining methods have two problems as follows. 1) In the hardness measurement, the important structural information is overlooked for similarity calculation, degrading the representativeness of the selected hard negative samples. 2) Previous works merely focus on the hard negative sample pairs while neglecting the hard positive sample pairs. Nevertheless, samples within the same cluster but with low similarity should also be carefully learned. To solve the problems, we propose a novel contrastive deep graph clustering method dubbed Hard Sample Aware Network (HSAN) by introducing a comprehensive similarity measure criterion and a general dynamic sample weighing strategy. Concretely, in our algorithm, the similarities between samples are calculated by considering both the attribute embeddings and the structure embeddings, better revealing sample relationships and assisting hardness measurement. Moreover, under the guidance of the carefully collected high-confidence clustering information, our proposed weight modulating function will first recognize the positive and negative samples and then dynamically up-weight the hard sample pairs while down-weighting the easy ones. In this way, our method can mine not only the hard negative samples but also the hard positive sample, thus improving the discriminative capability of the samples further. Extensive experiments and analyses demonstrate the superiority and effectiveness of our proposed method.

cs.LG↗

arXiv2023

Meta-Learning PAC-Bayes Priors in Model Averaging

Nowadays model uncertainty has become one of the most important problems in both academia and industry. In this paper, we mainly consider the scenario in which we have a common model set used for model averaging instead of selecting a single final model via a model selection procedure to account for this model's uncertainty to improve the reliability and accuracy of inferences. Here one main challenge is to learn the prior over the model set. To tackle this problem, we propose two data-based algorithms to get proper priors for model averaging. One is for meta-learner, the analysts should use historical similar tasks to extract the information about the prior. The other one is for base-learner, a subsampling method is used to deal with the data step by step. Theoretically, an upper bound of risk for our algorithm is presented to guarantee the performance of the worst situation. In practice, both methods perform well in simulations and real data studies, especially with poor-quality data.

cs.LG↗

arXiv2023

The phonon thermal Hall angle in black phosphorus

The origin of phonon thermal Hall Effect (THE) observed in a variety of insulators is yet to be identified. Here, we report on the observation of a thermal Hall conductivity in a non-magnetic elemental insulator, with an amplitude exceeding what has been previously observed. In black phosphorus (BP), the longitudinal ($κ_{ii}$), and the transverse, $κ_{ij}$, thermal conductivities peak at the same temperature and at this peak temperature, the $κ_{ij}/κ_{jj}/B $ is $\approx 10^{-4}$-$10^{-3}$ T$^{-1}$. Both these features are shared by other insulators displaying THE, despite an absolute amplitude spreading over three orders of magnitude. The absence of correlation between the thermal Hall angle and the phonon mean-free-path imposes a severe constraint for theoretical scenarios of THE. We show that in BP a longitudinal and a transverse acoustic phonon mode anti-cross, facilitating wave-like transport across modes and the anisotropic charge distribution surrounding atomic bonds, paving the way for coupling with magnetic field.

cond-mat.str-el↗

arXiv2022

Classification of radial non-radiative solutions to the 5D nonlinear wave equations

In this work we classify all radial non-radiative solutions to the 5D nonlinear wave equations with a wide range of energy critical nonlinearity. We show that such a solution always comes with two characteristic numbers. These characteristic numbers can be determined by either the radiation profile of the initial data or the asymptotic behaviour of the solution. In addition, two radial weakly non-radiative solutions with the same characteristic numbers must coincide with each other in the overlap part of their exterior regions. Finally we give a few applications of our theory on the global behaviours of solutions to the nonlinear wave equations.

math.AP↗

arXiv2022

BEV-Locator: An End-to-end Visual Semantic Localization Network Using Multi-View Images

Accurate localization ability is fundamental in autonomous driving. Traditional visual localization frameworks approach the semantic map-matching problem with geometric models, which rely on complex parameter tuning and thus hinder large-scale deployment. In this paper, we propose BEV-Locator: an end-to-end visual semantic localization neural network using multi-view camera images. Specifically, a visual BEV (Birds-Eye-View) encoder extracts and flattens the multi-view images into BEV space. While the semantic map features are structurally embedded as map queries sequence. Then a cross-model transformer associates the BEV features and semantic map queries. The localization information of ego-car is recursively queried out by cross-attention modules. Finally, the ego pose can be inferred by decoding the transformer outputs. We evaluate the proposed method in large-scale nuScenes and Qcraft datasets. The experimental results show that the BEV-locator is capable to estimate the vehicle poses under versatile scenarios, which effectively associates the cross-model information from multi-view images and global semantic maps. The experiments report satisfactory accuracy with mean absolute errors of 0.052m, 0.135m and 0.251$^\circ$ in lateral, longitudinal translation and heading angle degree.

cs.CV↗

arXiv2022

Cross-Architecture Knowledge Distillation

Transformer attracts much attention because of its ability to learn global relations and superior performance. In order to achieve higher performance, it is natural to distill complementary knowledge from Transformer to convolutional neural network (CNN). However, most existing knowledge distillation methods only consider homologous-architecture distillation, such as distilling knowledge from CNN to CNN. They may not be suitable when applying to cross-architecture scenarios, such as from Transformer to CNN. To deal with this problem, a novel cross-architecture knowledge distillation method is proposed. Specifically, instead of directly mimicking output/intermediate features of the teacher, partially cross attention projector and group-wise linear projector are introduced to align the student features with the teacher's in two projected feature spaces. And a multi-view robust training scheme is further presented to improve the robustness and stability of the framework. Extensive experiments show that the proposed method outperforms 14 state-of-the-arts on both small-scale and large-scale datasets.

cs.CV↗

arXiv2022

Fast-R2D2: A Pretrained Recursive Neural Network based on Pruned CKY for Grammar Induction and Text Representation

Recently CKY-based models show great potential in unsupervised grammar induction thanks to their human-like encoding paradigm, which runs recursively and hierarchically, but requires $O(n^3)$ time-complexity. Recursive Transformer based on Differentiable Trees (R2D2) makes it possible to scale to large language model pre-training even with complex tree encoder by introducing a heuristic pruning method. However, the rule-based pruning approach suffers from local optimum and slow inference issues. In this paper, we fix those issues in a unified method. We propose to use a top-down parser as a model-based pruning method, which also enables parallel encoding during inference. Typically, our parser casts parsing as a split point scoring task, which first scores all split points for a given sentence, and then recursively splits a span into two by picking a split point with the highest score in the current span. The reverse order of the splits is considered as the order of pruning in R2D2 encoder. Beside the bi-directional language model loss, we also optimize the parser by minimizing the KL distance between tree probabilities from parser and R2D2. Our experiments show that our Fast-R2D2 improves performance significantly in grammar induction and achieves competitive results in downstream classification tasks.

cs.CL↗

arXiv2022

On the X-ray, optical and radio afterglows of the BdHN I GRB 180720B generated by the synchrotron emission

Gamma-ray bursts (GRBs) are systems of unprecedented complexity across all the electromagnetic spectrum, including the radio, optical, X-rays, gamma-rays in the megaelectronvolt (MeV) and gigaelectronvolt (GeV) regime, as well as ultrahigh-energy cosmic rays (UHECRs), each manifested in seven specific physical processes with widely different characteristic evolution timescales ranging from $10^{-14}$ s to $10^{7}$ s or longer. We here study the long GRB 180720B originating from a binary system composed of a massive carbon-oxygen (CO) star of about $10 M_\odot$ and a companion neutron star (NS). The gravitational collapse of the CO star gives rise to a spinning newborn NS ($ν$NS), with an initial period of $P_0=1$ ms that powers the synchrotron radiation in the radio, optical, and X-ray wavelengths. We here only investigate the GRB 180720B afterglows and present a detailed treatment of its origin based on the synchrotron radiation released by the interaction of the $ν$NS and the SN ejecta. We show that in parallel to the X-ray afterglow, the spinning $ν$NS also powers the optical and radio afterglows and allows us to infer the $ν$NS and ejecta parameters that fit the observational data.

astro-ph.HE↗

arXiv2022

Graph-to-Text Generation with Dynamic Structure Pruning

Most graph-to-text works are built on the encoder-decoder framework with cross-attention mechanism. Recent studies have shown that explicitly modeling the input graph structure can significantly improve the performance. However, the vanilla structural encoder cannot capture all specialized information in a single forward pass for all decoding steps, resulting in inaccurate semantic representations. Meanwhile, the input graph is flatted as an unordered sequence in the cross attention, ignoring the original graph structure. As a result, the obtained input graph context vector in the decoder may be flawed. To address these issues, we propose a Structure-Aware Cross-Attention (SACA) mechanism to re-encode the input graph representation conditioning on the newly generated context at each decoding step in a structure aware manner. We further adapt SACA and introduce its variant Dynamic Graph Pruning (DGP) mechanism to dynamically drop irrelevant nodes in the decoding process. We achieve new state-of-the-art results on two graph-to-text datasets, LDC2020T02 and ENT-DESC, with only minor increase on computational cost.

cs.CL↗

arXiv2022

YOLOv6: A Single-Stage Object Detection Framework for Industrial Applications

For years, the YOLO series has been the de facto industry-level standard for efficient object detection. The YOLO community has prospered overwhelmingly to enrich its use in a multitude of hardware platforms and abundant scenarios. In this technical report, we strive to push its limits to the next level, stepping forward with an unwavering mindset for industry application. Considering the diverse requirements for speed and accuracy in the real environment, we extensively examine the up-to-date object detection advancements either from industry or academia. Specifically, we heavily assimilate ideas from recent network design, training strategies, testing techniques, quantization, and optimization methods. On top of this, we integrate our thoughts and practice to build a suite of deployment-ready networks at various scales to accommodate diversified use cases. With the generous permission of YOLO authors, we name it YOLOv6. We also express our warm welcome to users and contributors for further enhancement. For a glimpse of performance, our YOLOv6-N hits 35.9% AP on the COCO dataset at a throughput of 1234 FPS on an NVIDIA Tesla T4 GPU. YOLOv6-S strikes 43.5% AP at 495 FPS, outperforming other mainstream detectors at the same scale~(YOLOv5-S, YOLOX-S, and PPYOLOE-S). Our quantized version of YOLOv6-S even brings a new state-of-the-art 43.3% AP at 869 FPS. Furthermore, YOLOv6-M/L also achieves better accuracy performance (i.e., 49.5%/52.3%) than other detectors with a similar inference speed. We carefully conducted experiments to validate the effectiveness of each component. Our code is made available at https://github.com/meituan/YOLOv6.

cs.CV↗

arXiv2022

EpiGNN: Exploring Spatial Transmission with Graph Neural Network for Regional Epidemic Forecasting

Epidemic forecasting is the key to effective control of epidemic transmission and helps the world mitigate the crisis that threatens public health. To better understand the transmission and evolution of epidemics, we propose EpiGNN, a graph neural network-based model for epidemic forecasting. Specifically, we design a transmission risk encoding module to characterize local and global spatial effects of regions in epidemic processes and incorporate them into the model. Meanwhile, we develop a Region-Aware Graph Learner (RAGL) that takes transmission risk, geographical dependencies, and temporal information into account to better explore spatial-temporal dependencies and makes regions aware of related regions' epidemic situations. The RAGL can also combine with external resources, such as human mobility, to further improve prediction performance. Comprehensive experiments on five real-world epidemic-related datasets (including influenza and COVID-19) demonstrate the effectiveness of our proposed method and show that EpiGNN outperforms state-of-the-art baselines by 9.48% in RMSE.

q-bio.QM↗

arXiv2022

The structure of the ultrarelativistic prompt emission phase and the properties of the black hole in GRB 180720B

In analogy with GRB 190114C, we here analyze the ultrarelativistic prompt emission (UPE) of GRB 180720B observed in the rest-frame time interval $t_{\rm rf}=4.84$--$10.89$~s by Fermi-GBM. We reveal the UPE hierarchical structure from the time-resolved spectral analysis performed in time sub-intervals: the spectrum in each shorter time interval is always fitted by a composite blackbody plus cutoff power-law model. We explain this structure with the \textit{inner engine} of binary-driven hypernova (BdHN) model operating in a quantum electrodynamics (QED) regime. In this regime, the electric field induced by the gravitomagnetic interaction of the newborn Kerr BH with the surrounding magnetic field is overcritical, i.e., $|{\bf E}|\geq E_c$, where $E_c=m_e^2 c^3/(e\hbar)$. The overcritical field polarizes the vacuum leading to an $e^+~e^-$ pair plasma that loads baryons from the surroundings during its expansion. We calculate the dynamics of the self-acceleration of the pair-electromagnetic-baryon (PEMB) pulses to their point of transparency. We characterize the quantum vacuum polarization process in the sequences of decreasing time bins of the UPE by determining the radiation timescale, Lorentz factors, and transparency radius of the PEMB pulses. We also estimate the strength of the surrounding magnetic field $\sim 10^{14}$ G, and obtain a lower limit to the BH mass, $M=2.4~M_\odot$, and correspondingly an upper limit to the spin, $α= 0.6$, from the conditions that the UPE is powered by the Kerr BH extractable energy and its mass is bound from below by the NS critical mass.

astro-ph.HE↗