arXiv subjects

Jing Liu

Publications and source records attributed to Jing Liu.

At least 523 records · Page 29Linked to original sources

arXiv2022

New analysis of the fraction of observable nights at astronomical sites based on FengYun-2 satellite data

The fraction of observable nights is an essential parameter for selecting astronomical sites. In recent years, meteorological satellite data have played an essential role in recognising and providing statistics of observable nights. We present a method to estimate the fraction of observable nights based on the FengYun-2 series of geostationary meteorological satellites and weather records of multiple astronomical sites. We have calculated the fraction of observable nights at 27 sites in Indonesia and two astronomical sites in China to validate the method. The results derived from our method show good agreement with previous works. Furthermore, we have derived the yearly distribution of the fraction of observable nights above China, which indicates the area near 40$^{\circ}$N has more observable nights than other areas in China.

astro-ph.IM↗

arXiv2022

First-principles perspective on full-spectrum infrared photodetectors from doping an excitonic insulator

Innovations in imaging technology involves finding strategies and materials suitable for detection applications over the entire infrared range. Herein, we propose a new design concept based on the unique feature of an excitonic insulator, namely, negative exciton transition energy ($E_t$). We demonstrate this concept using first-principles $GW$-BSE calculations on one-dimensional organometallic wire (CrBz)$_\infty$. The pristine (CrBz)$_\infty$ exhibits an excitonic instability due to a negative $E_t$ for the lowest exciton. Substitutional doping can continuously tune the $E_t$ from $\sim$0 to $\sim$0.6 eV, which shows the ability of photon detection from terahertz to near-infrared. This type of detectors have advantages of outstanding wavelength selectivity, reduced thermal disturbance and elevated working temperature. Our work not only adds another member in the family of rare one-dimensional excitonic insulators, but also opens a new avenue for the development of high-performance infrared photodetectors in the future.

cond-mat.mtrl-sci↗

arXiv2022

Primordial black hole production during first-order phase transitions

Primordial black holes (PBHs) produced in the early Universe have attracted wide interest for their ability to constitute dark matter and explain the compact binary coalescence. We propose a new mechanism of PBH production during first-order phase transitions (PTs) and find that PBHs are naturally produced during PTs model-independently. Because of the randomness of the quantum tunneling, there always exists some probability that the vacuum decay is postponed in a whole Hubble volume. Since the vacuum energy density remains constant while radiation is quickly redshifted in the expanding Universe, the postponed vacuum decay then results in overdense regions, which finally collapse into PBHs as indicated by numerical simulations. Utilizing this result one can obtain mutual predictions and constraints between PBHs and GWs from PTs. The predicted mass function of PBHs is nearly monochromatic. We investigate two typical cases and find that 1) PBHs from a PT constitute all dark matter and GWs peak at $1$Hz, 2) PBHs from a PT can explain the coalescence events observed by LIGO-Virgo collaboration, and meanwhile GWs can explain the common-spectrum process detected by NANOGrav collaboration.

astro-ph.CO↗

arXiv2022

Dependence of the amplitude of gravitational waves from preheating on the inflationary energy scale

Stochastic gravitational wave backgrounds (SGWBs) receive increasing attention and provide a new possibility to directly probe the early Universe. In the preheating process at the end of inflation, parametric resonance can generate large energy density perturbations and efficiently produce gravitational waves (GWs) which carry unique information about inflation. Since the peak frequency of such GWs is approximately proportional to the inflationary energy scale, $Λ_{\mathrm{inf}}$, GWs from preheating are expected to be observed by interferometer GW detectors in low-scale inflationary models. We investigate the dependence of the amplitude of such GWs on $Λ_{\mathrm{inf}}$, and find that the present energy spectrum of these GWs does not depend on $Λ_{\mathrm{inf}}$ only in the case of $Λ_{\mathrm{inf}}$ is above a critical value $Λ_{c}$, a parameter depending on the resonance strength. We numerically obtain $Λ_{c}$ in terms of the model parameters in linear approximation and then conduct lattice simulations to verify this result. For $Λ_{\mathrm{inf}}\lesssimΛ_{c}$, the amplitude of GWs quickly decreases with $Λ_{\mathrm{inf}}$ and becomes challenging to observe. In turn, observing such GWs in interferometer detectors also helps to determine $Λ_{\mathrm{inf}}$ and the resonance strength during the preheating.

astro-ph.CO↗

arXiv2022

Evolutionary Multitasking AUC Optimization

Learning to optimize the area under the receiver operating characteristics curve (AUC) performance for imbalanced data has attracted much attention in recent years. Although there have been several methods of AUC optimization, scaling up AUC optimization is still an open issue due to its pairwise learning style. Maximizing AUC in the large-scale dataset can be considered as a non-convex and expensive problem. Inspired by the characteristic of pairwise learning, the cheap AUC optimization task with a small-scale dataset sampled from the large-scale dataset is constructed to promote the AUC accuracy of the original, large-scale, and expensive AUC optimization task. This paper develops an evolutionary multitasking framework (termed EMTAUC) to make full use of information among the constructed cheap and expensive tasks to obtain higher performance. In EMTAUC, one mission is to optimize AUC from the sampled dataset, and the other is to maximize AUC from the original dataset. Moreover, due to the cheap task containing limited knowledge, a strategy for dynamically adjusting the data structure of inexpensive tasks is proposed to introduce more knowledge into the multitasking AUC optimization environment. The performance of the proposed method is evaluated on a series of binary classification datasets. The experimental results demonstrate that EMTAUC is highly competitive to single task methods and online methods. Supplementary materials and source code implementation of EMTAUC can be accessed at https://github.com/xiaofangxd/EMTAUC.

cs.LG↗

arXiv2021

Less is More: Pay Less Attention in Vision Transformers

Transformers have become one of the dominant architectures in deep learning, particularly as a powerful alternative to convolutional neural networks (CNNs) in computer vision. However, Transformer training and inference in previous works can be prohibitively expensive due to the quadratic complexity of self-attention over a long sequence of representations, especially for high-resolution dense prediction tasks. To this end, we present a novel Less attention vIsion Transformer (LIT), building upon the fact that the early self-attention layers in Transformers still focus on local patterns and bring minor benefits in recent hierarchical vision Transformers. Specifically, we propose a hierarchical Transformer where we use pure multi-layer perceptrons (MLPs) to encode rich local patterns in the early stages while applying self-attention modules to capture longer dependencies in deeper layers. Moreover, we further propose a learned deformable token merging module to adaptively fuse informative patches in a non-uniform manner. The proposed LIT achieves promising performance on image recognition tasks, including image classification, object detection and instance segmentation, serving as a strong backbone for many vision tasks. Code is available at: https://github.com/zhuang-group/LIT

cs.CV↗

arXiv2021

Proof of Dilks' bijectivity conjecture on Baxter permutations

Baxter permutations originally arose in studying common fixed points of two commuting continuous functions. In 2015, Dilks proposed a conjectured bijection between Baxter permutations and non-intersecting triples of lattice paths in terms of inverse descent bottoms, descent positions and inverse descent tops. We prove this bijectivity conjecture by investigating its connection with the Françon--Viennot bijection. As a result, we obtain a permutation interpretation of the $(t,q)$-analog of the Baxter numbers $$ \frac{1}{{n+1\brack 1}_q{n+1\brack 2}_q}\sum_{k=0}^{n-1}q^{3{k+1\choose2}}{n+1\brack k}_q{n+1\brack k+1}_q{n+1\brack k+2}_qt^k, $$ where ${n\brack k}_q$ denote the $q$-binomial coefficients.

math.CO↗

arXiv2021

A New Picture of Cosmic String Evolution and Anisotropic Stochastic Gravitational-Wave Background

We investigate the anisotropies of the stochastic gravitational-wave background (SGWB) produced by cosmic strings associated with the spontaneous U(1) symmetry breaking of Grand Unified Theory, which happens at the onset of inflation. The string network evolution is determined by primordial fluctuations and never reaches the scaling regime. The string loops are inhomogeneously distributed in large scale regions, resulting in large anisotropies in the SGWB. We find that the angular power spectrum of SGWB anisotropies depends on frequency, which is testable in multiband observations of GWs. In particular, GWs from the cosmic strings can appropriately interpret the common-spectrum process reported by NANOGrav collaboration, and the angular power spectrum in the nanohertz band, $\mathtt{l}(\mathtt{l}+1)C_{\mathtt{l}}=5.6\times 10^{-2}$ at large scales, is expected to be detectable by pulsar timing array experiments in the near future.

astro-ph.CO↗

arXiv2021

Attentive Contextual Carryover for Multi-Turn End-to-End Spoken Language Understanding

Recent years have seen significant advances in end-to-end (E2E) spoken language understanding (SLU) systems, which directly predict intents and slots from spoken audio. While dialogue history has been exploited to improve conventional text-based natural language understanding systems, current E2E SLU approaches have not yet incorporated such critical contextual signals in multi-turn and task-oriented dialogues. In this work, we propose a contextual E2E SLU model architecture that uses a multi-head attention mechanism over encoded previous utterances and dialogue acts (actions taken by the voice assistant) of a multi-turn dialogue. We detail alternative methods to integrate these contexts into the state-ofthe-art recurrent and transformer-based models. When applied to a large de-identified dataset of utterances collected by a voice assistant, our method reduces average word and semantic error rates by 10.8% and 12.6%, respectively. We also present results on a publicly available dataset and show that our method significantly improves performance over a noncontextual baseline

cs.CL↗

arXiv2021

Bhatia-Davis formula in the quantum speed limit

The Bhatia-Davis theorem provides a useful upper bound for the variance in mathematics, and in quantum mechanics, the variance of a Hamiltonian is naturally connected to the quantum speed limit due to the Mandelstam-Tamm bound. Inspired by this connection, we construct a formula, referred to as the Bhatia-Davis formula, for the characterization of the quantum speed limit in the Bloch representation. We first prove that the Bhatia-Davis formula is an upper bound for a recently proposed operational definition of the quantum speed limit, which means it can be used to reveal the closeness between the timescale of certain chosen states to the systematic minimum timescale. In the case of the largest target angle, the Bhatia-Davis formula is proved to be a valid lower bound for the evolution time to reach the target when the energy structure is symmetric. Regarding few-level systems, it is also proved to be a valid lower bound for any state in two-level systems with any target, and for most mixed states with large target angles in equally spaced three-level systems.

quant-ph↗

arXiv2021

Optimal Scheme for Quantum Metrology

Quantum metrology can achieve far better precision than classical metrology, and is one of the most important applications of quantum technologies in the real world. To attain the highest precision promised by quantum metrology, all steps of the schemes need to be optimized, which include the state preparation, parametrization, and measurement. Here the recent progresses on the optimization of these steps, which are essential for the identification and achievement of the ultimate precision limit in quantum metrology, are reviewed. It is hoped this provides a useful reference for the researchers in quantum metrology and related fields.

quant-ph↗

arXiv2021

AutoCaption: Image Captioning with Neural Architecture Search

Image captioning transforms complex visual information into abstract natural language for representation, which can help computers understanding the world quickly. However, due to the complexity of the real environment, it needs to identify key objects and realize their connections, and further generate natural language. The whole process involves a visual understanding module and a language generation module, which brings more challenges to the design of deep neural networks than other tasks. Neural Architecture Search (NAS) has shown its important role in a variety of image recognition tasks. Besides, RNN plays an essential role in the image captioning task. We introduce a AutoCaption method to better design the decoder module of the image captioning where we use the NAS to design the decoder module called AutoRNN automatically. We use the reinforcement learning method based on shared parameters for automatic design the AutoRNN efficiently. The search space of the AutoCaption includes connections between the layers and the operations in layers both, and it can make AutoRNN express more architectures. In particular, RNN is equivalent to a subset of our search space. Experiments on the MSCOCO datasets show that our AutoCaption model can achieve better performance than traditional hand-design methods.

cs.CV↗

arXiv2021

Single-photon-triggered spin squeezing with decoherence reduction in optomechanics via phase matching

Quantum spin squeezing is an important resource for quantum information processing, but its squeezing degree is not easy to preserve in an open system with decoherence. Here, we propose a scheme to implement single-photon-triggered spin squeezing with decoherence reduction in an open system. In our system, a Dicke model (DM) is introduced into the quadratic optomechanics, which can be equivalent to an effective DM manipulated by the photon number. Besides, the phonon mode of the optomechanical system is coupled to a squeezed vacuum reservoir with a phase matching, resulting in that the thermal noise caused by the environment can be suppressed completely. We show that squeezing of the phonon mode triggered by a single photon can be transferred to the spin ensemble totally, and pairwise entanglement of the spin ensemble can be realized if and only if there is spin squeezing. Importantly, when considering the impact of the environment, our system can obtain a better squeezing degree than the optimal squeezing that can be achieved in the traditional DM. Meanwhile, the spin squeezing generated in our system is immune to the thermal noise. This work offers an effective way to generate spin squeezing with a single photon and to reduce decoherence in an open system, which will have promising applications in quantum information processing.

quant-ph↗

arXiv2021

Context-Aware Transformer Transducer for Speech Recognition

End-to-end (E2E) automatic speech recognition (ASR) systems often have difficulty recognizing uncommon words, that appear infrequently in the training data. One promising method, to improve the recognition accuracy on such rare words, is to latch onto personalized/contextual information at inference. In this work, we present a novel context-aware transformer transducer (CATT) network that improves the state-of-the-art transformer-based ASR system by taking advantage of such contextual signals. Specifically, we propose a multi-head attention-based context-biasing network, which is jointly trained with the rest of the ASR sub-networks. We explore different techniques to encode contextual data and to create the final attention context vectors. We also leverage both BLSTM and pretrained BERT based models to encode contextual data and guide the network training. Using an in-house far-field dataset, we show that CATT, using a BERT based context encoder, improves the word error rate of the baseline transformer transducer and outperforms an existing deep contextual model by 24.2% and 19.4% respectively.

cs.CL↗

arXiv2021

The Gravitational-Wave Physics II: Progress

It has been a half-decade since the first direct detection of gravitational waves, which signifies the coming of the era of the gravitational-wave astronomy and gravitational-wave cosmology. The increasing number of the detected gravitational-wave events has revealed the promising capability of constraining various aspects of cosmology, astronomy, and gravity. Due to the limited space in this review article, we will briefly summarize the recent progress over the past five years, but with a special focus on some of our own work for the Key Project ``Physics associated with the gravitational waves'' supported by the National Natural Science Foundation of China. In particular, (1) we have presented the mechanism of the gravitational-wave production during some physical processes of the early Universe, such as inflation, preheating and phase transition, and the cosmological implications of gravitational-wave measurements; (2) we have put constraints on the neutron star maximum mass according to GW170817 observations; (3) we have developed a numerical relativity algorithm based on the finite element method and a waveform model for the binary black hole coalescence along an eccentric orbit.

gr-qc↗

arXiv2021

Partial Hadamard Encoded Synthetic Transmit Aperture for High Frame Rate Imaging with Minimal l2-Norm Least Square Method

Synthetic transmit aperture (STA) ultrasound imaging is well known for ideal focusing in the full field of view. However, it suffers from low signal-to-noise ratio (SNR) and low frame rate, because each array element must be activated individually. In our previous study, we encoded all the array elements with partial Hadamard matrix and reconstructed the complete STA dataset with compressed sensing (CS) algorithm (CS-STA). As all the elements are activated in each transmission and the number of transmissions is smaller than that of STA, this method can achieve higher SNR and higher frame rate. Its main drawback is the time-consuming CS reconstruction. In this study, we accelerate the complete STA dataset reconstruction with minimal l2-norm least square method. Thanks of the orthogonality of partial Hadamard matrix, the minimal l2-norm least square solution can be easily calculated. The proposed method is tested with simulation data and experimental phantom and in-vivo data. The results demonstrate that the proposed method achieves ~5*10^3 times faster reconstruction speed than CS algorithm. The simulation results demonstrate that the proposed method is capable of achieving the same accuracy for STA dataset reconstruction as conventional CS-STA method. The simulations, phantom and in-vivo experiments show that the proposed method is capable of improving the generalized contrast-to-noise ratio (gCNR) and SNR with maintained spatial resolution and fewer transmissions, compared with STA. In conclusion, the improved image quality and reduced computational time of LS-STA pave the way for its real-time applications in the clinics.

physics.med-ph↗

arXiv2021

Gravitational waves from resonant amplification of curvature perturbations during inflation

Parametric resonance in a single-field inflationary model with a periodic structure on the potential gives rise to curvature perturbations with large amplitudes on small scales, which could result in observable primordial black holes (PBHs) and concomitant gravitational waves (GWs) induced by curvature perturbations in the radiation-dominated era. In such a model, GWs associated with the PBH formation were investigated in Ref. [1]. In this paper, we consider a stochastic GW background sourced by inflaton perturbations resonantly amplified during inflation. We compute the energy spectra of induced GWs produced both during inflation and in the radiation-dominated era, and find that the peak of the energy spectrum of the former is much higher than that of the latter, but is located at a lower frequency. Moreover, the energy spectrum of induced GWs produced during inflation exhibits a unique oscillating character in the ultraviolet region. Both the stochastic GW backgrounds are expected to be detected by future space-based laser interferometers.

astro-ph.CO↗

arXiv2021

Multi-local Collaborative AutoEncoder

The excellent performance of representation learning of autoencoders have attracted considerable interest in various applications. However, the structure and multi-local collaborative relationships of unlabeled data are ignored in their encoding procedure that limits the capability of feature extraction. This paper presents a Multi-local Collaborative AutoEncoder (MC-AE), which consists of novel multi-local collaborative representation RBM (mcrRBM) and multi-local collaborative representation GRBM (mcrGRBM) models. Here, the Locality Sensitive Hashing (LSH) method is used to divide the input data into multi-local cross blocks which contains multi-local collaborative relationships of the unlabeled data and features since the similar multi-local instances and features of the input data are divided into the same block. In mcrRBM and mcrGRBM models, the structure and multi-local collaborative relationships of unlabeled data are integrated into their encoding procedure. Then, the local hidden features converges on the center of each local collaborative block. Under the collaborative joint influence of each local block, the proposed MC-AE has powerful capability of representation learning for unsupervised clustering. However, our MC-AE model perhaps perform training process for a long time on the large-scale and high-dimensional datasets because more local collaborative blocks are integrate into it. Five most related deep models are compared with our MC-AE. The experimental results show that the proposed MC-AE has more excellent capabilities of collaborative representation and generalization than the contrastive deep models.

cs.LG↗