Search

Recently, many works have designed wider and deeper networks to achieve higher image super-resolution performance. Despite their outstanding performance, they still suffer from high computational resources, preventing them from directly applying to embedded devices. To reduce the computation resources and maintain performance, we propose a novel Ghost Residual Attention Network (GRAN) for efficient super-resolution. This paper introduces Ghost Residual Attention Block (GRAB) groups to overcome the drawbacks of the standard convolutional operation, i.e., redundancy of the intermediate feature. GRAB consists of the Ghost Module and Channel and Spatial Attention Module (CSAM) to alleviate the generation of redundant features. Specifically, Ghost Module can reveal information underlying intrinsic features by employing linear operations to replace the standard convolutions. Reducing redundant features by the Ghost Module, our model decreases memory and computing resource requirements in the network. The CSAM pays more comprehensive attention to where and what the feature extraction is, which is critical to recovering the image details. Experiments conducted on the benchmark datasets demonstrate the superior performance of our method in both qualitative and quantitative. Compared to the baseline models, we achieve higher performance with lower computational resources, whose parameters and FLOPs have decreased by more than ten times.

The nonanalyticity of the Loschmidt echo at critical times in quantum quenched systems is termed as the dynamical quantum phase transition, extending the notion of quantum criticality to a nonequilibrium scenario. In this paper, we establish a new paradigm of dynamical phase transitions driven by a sudden change in the internal spatial correlations of the disorder potential in a low-dimensional disordered system. The quench dynamics between prequenched pure and postquenched random system Hamiltonian reveals an anomalous quantum dynamical quantum phase transition triggered by an infinite disorder correlation in the modulation potential. The physical origin of the anomalous phenomenon is associated with the overlap between the two distinctly different extended states. Furthermore, we explore the quench dynamics between the prequenched random and postquenched pure system Hamiltonian. Interestingly, the quenched system undergoes dynamical quantum phase transitions for the prequench white-noise potential in the thermodynamic limit. In addition, the quench dynamics also shows a clear signature of the delocalization phase transition in the correlated Anderson model., Comment: 9 pages, 9 figures

Recovering clear structures from severely blurry inputs is a challenging problem due to the large movements between the camera and the scene. Although some works apply segmentation maps on human face images for deblurring, they cannot handle natural scenes because objects and degradation are more complex, and inaccurate segmentation maps lead to a loss of details. For general scene deblurring, the feature space of the blurry image and corresponding sharp image under the high-level vision task is closer, which inspires us to rely on other tasks (e.g. classification) to learn a comprehensive prior in severe blur removal cases. We propose a cross-level feature learning strategy based on knowledge distillation to learn the priors, which include global contexts and sharp local structures for recovering potential details. In addition, we propose a semantic prior embedding layer with multi-level aggregation and semantic attention transformation to integrate the priors effectively. We introduce the proposed priors to various models, including the UNet and other mainstream deblurring baselines, leading to better performance on severe blur removal. Extensive experiments on natural image deblurring benchmarks and real-world images, such as GoPro and RealBlur datasets, demonstrate our method's effectiveness and generalization ability.

We solve the fully-connected Ising model in the presence of dissipation and time-periodic field, with the corresponding Lindblad equation having a time-periodic Liouvillian. The dynamics of the magnetizations is studied by using both the semiclassical approach and the numerical simulation with the help of permutation symmetry. The semiclassical approach shows a transition from the periodic response for small field amplitude to the chaotic dynamics for large amplitude. The trajectory of the magnetizations and the Lyapunov exponents are calculated, which support the existence of a chaotic phase. But in the exact numerical simulation, the response is periodic for both small and large amplitude. The scaling analysis of Floquet Liouvillian spectrum confirms the periodic response in the thermodynamic limit. The semiclassical approximation is found to fail as the field amplitude is large., Comment: 8 pages, 8 figures

Fast radio bursts (FRBs) are highly dispersed millisecond-duration radio bursts, of which the physical origin is still not fully understood. FRB 20201124A is one of the most actively repeating FRBs. In this paper, we present the collection of 1863 burst dynamic spectra of FRB 20201124A measured with the Five-hundred-meter Aperture Spherical radio Telescope (FAST). The current collection, taken from the observation during the FRB active phase from April to June 2021, is the largest burst sample detected in any FRB so far. The standard PSRFITs format is adopted, including dynamic spectra of the burst, and the time information of the dynamic spectra, in addition, mask files help readers to identify the pulse positions are also provided.

Fast radio bursts (FRBs) are bright millisecond radio bursts at cosmological distances. Only a small fraction of FRBs apparently repeat. Polarization, a fundamental property of electromagnetic signals, often carries critical information about the radiation processes, the environment, and the intervening medium of FRBs. Here we report circular polarization detections of two active repeating FRBs, namely FRBs 20121102A and 20190520B, with the Five-hundred-meter Aperture Spherical radio Telescope. We detect circular polarization in both active repeating FRBs, which increases the number of repeating FRBs with circular polarization to three. In one of the bursts of FRB 20121102A, we detect 64% degree of circular polarization. The observed circular polarization is unlikely induced by multipath propagation. Our observations favor circular polarization induced by Faraday conversion or radiation mechanism intrinsic to the FRB source. The conditions to generate circular polarization have to be rare in either case., Comment: 9pages, 3 figures, and 1 table, published in Science Bulletin

Fast radio bursts (FRBs) are short pulses observed in radio frequencies usually originating from cosmological distances. The discovery of FRB 200428 and its X-ray counterpart from the Galactic magnetar SGR J1935+2154 suggests that at least some FRBs can be generated by magnetars. However, the majority of X-ray bursts from magnetars are not associated with radio emission. The fact that only in rare cases can an FRB be generated raises the question regarding the special triggering mechanism of FRBs. Here we report a giant glitch from SGR J1935+2154, which occurred approximately $3.1\pm2.5$\,day before FRB 200428, with $\Delta\nu=19.8\pm1.4$ {\rm $\mu$Hz} and $\Delta\dot{\nu}=6.3\pm1.1$\,pHz s$^{-1}$. The corresponding spin-down power change rate $\Delta\dot\nu/\dot\nu$ is among the largest in all the detected pulsar glitches. The glitch contains a delayed spin-up process that is only detected in the Crab pulsar and the magnetar 1E 2259+586, a large persistent offset of the spin-down rate, and a recovery component which is about one order of magnitude smaller than the persistent one. The temporal coincidence between the glitch and FRB 200428 suggests a physical connection between the two. The internally triggered giant glitch of the magnetar likely altered the magnetosphere structure dramatically in favour of FRB generation, which subsequently triggered many X-ray bursts and eventually FRB 200428 through additional crustal cracking and Alfv\'en wave excitation and propagation.

Detecting out-of-domain (OOD) intents from user queries is essential for avoiding wrong operations in task-oriented dialogue systems. The key challenge is how to distinguish in-domain (IND) and OOD intents. Previous methods ignore the alignment between representation learning and scoring function, limiting the OOD detection performance. In this paper, we propose a unified neighborhood learning framework (UniNL) to detect OOD intents. Specifically, we design a K-nearest neighbor contrastive learning (KNCL) objective for representation learning and introduce a KNN-based scoring function for OOD detection. We aim to align representation learning with scoring function. Experiments and analysis on two benchmark datasets show the effectiveness of our method., Comment: Accepted at EMNLP2022 main conference

Discovering out-of-domain (OOD) intent is important for developing new skills in task-oriented dialogue systems. The key challenges lie in how to transfer prior in-domain (IND) knowledge to OOD clustering, as well as jointly learn OOD representations and cluster assignments. Previous methods suffer from in-domain overfitting problem, and there is a natural gap between representation learning and clustering objectives. In this paper, we propose a unified K-nearest neighbor contrastive learning framework to discover OOD intents. Specifically, for IND pre-training stage, we propose a KCL objective to learn inter-class discriminative features, while maintaining intra-class diversity, which alleviates the in-domain overfitting problem. For OOD clustering stage, we propose a KCC method to form compact clusters by mining true hard negative samples, which bridges the gap between clustering and representation learning. Extensive experiments on three benchmark datasets show that our method achieves substantial improvements over the state-of-the-art methods., Comment: Accepted at EMNLP2022 main conference

Recent advances in neural approaches greatly improve task-oriented dialogue (TOD) systems which assist users to accomplish their goals. However, such systems rely on costly manually labeled dialogs which are not available in practical scenarios. In this paper, we present our models for Track 2 of the SereTOD 2022 challenge, which is the first challenge of building semi-supervised and reinforced TOD systems on a large-scale real-world Chinese TOD dataset MobileCS. We build a knowledge-grounded dialog model to formulate dialog history and local KB as input and predict the system response. And we perform semi-supervised pre-training both on the labeled and unlabeled data. Our system achieves the first place both in the automatic evaluation and human interaction, especially with higher BLEU (+7.64) and Success (+13.6\%) than the second place., Comment: Accepted at the SereTOD 2022 Workshop, EMNLP 2022

Detecting Out-of-Domain (OOD) or unknown intents from user queries is essential in a task-oriented dialog system. Traditional softmax-based confidence scores are susceptible to the overconfidence issue. In this paper, we propose a simple but strong energy-based score function to detect OOD where the energy scores of OOD samples are higher than IND samples. Further, given a small set of labeled OOD samples, we introduce an energy-based margin objective for supervised OOD detection to explicitly distinguish OOD samples from INDs. Comprehensive experiments and analysis prove our method helps disentangle confidence score distributions of IND and OOD data.\footnote{Our code is available at \url{https://github.com/pris-nlp/EMNLP2022-energy_for_OOD/}.}, Comment: accepted by the EMNLP2022 SereTOD workshop

A series of dynamical anomalies in the orbits of distant trans-Neptunian objects points to a new celestial body (usually named Planet Nine) in the solar system. In this draft, we point out that a mirror planet captured from the outer solar system or formed in the solar system is also a possible candidate. The introduction of the mirror matter model is due to an unbroken parity symmetry and is a potential explanation for dark matter. This mirror planet has null or fainter electromagnetic counterparts with a smaller optical radius and might be explored through gravitational effects.

Recent studies have revealed the global deposition on Earth of radioactive elements (e.g., $^{60}$Fe) resulting from the metal-enriched ejecta of nearby (within $\sim 100$ pc) supernova explosions. The majority of neutron stars in our Solar neighborhood remain to be discovered. Here we report the discovery of the nearest ($127.7 \pm 0.3$ pc) neutron star candidate in the single-lined spectroscopic binary LAMOST J235456.76+335625.7 (hereafter J2354). Utilizing the multi-epoch spectra and high-cadence periodic light curves, we measure the mass of the visible star ($M_{\rm vis}=0.70\pm 0.05\ M_{\odot}$) and determine the mass function of the invisible object $f(M)=0.525 \pm 0.004\ M_{\odot}$, i.e., the mass of the unseen compact object is $M_{\rm inv} \geq 1.26 \pm 0.03\ M_{\odot}$. The excess UV emission due to a hot supramassive white dwarf is absent. Hence, it is likely that J2354 harbors a neutron star. J2354 is X-ray dim (the $0.1$--$2.4$ keV luminosity $<10^{30}\ {\rm erg\ s^{-1}}$) since it is not detected in the ROSAT all-sky surveys in X-ray. One-hour exceptionally sensitive radio follow-up observations with FAST, the largest single-dish radio telescope, failed to reveal any radio pulsating signals (the potential pulse power at $1.4$ GHz is $<6.8\times 10^{23}\ {\rm erg\ s^{-1}}$). Hence, the neutron star candidate in J2354 can only be discovered via our time-resolved observations. The alternative scenario involving a nearby supramassive cold white dwarf cannot be fully excluded. Our discovery demonstrates a promising way to unveil the missing population of backyard inactive neutron stars or supramassive cold white dwarfs in binaries by exploring the optical time domain, thereby facilitating understanding of the supernovae explosion and metal-enrichment history in our Solar neighborhood., Comment: 35 pages, 8 figures, to be submitted

We report the properties of more than 800 bursts detected from the repeating fast radio burst (FRB) source FRB 20201124A with the Five-hundred-meter Aperture Spherical radio Telescope (FAST) during an extremely active episode on UTC September 25-28, 2021 in a series of four papers. In this second paper of the series, we mainly focus on the energy distribution of the detected bursts. The event rate initially increased exponentially but the source activity stopped within 24 hours after the 4th day. The detection of 542 bursts in one hour during the fourth day marked the highest event rate detected from one single FRB source so far. The bursts have complex structures in the time-frequency space. We find a double-peak distribution of the waiting time, which can be modeled with two log-normal functions peaking at 51.22 ms and 10.05 s, respectively. Compared with the emission from a previous active episode of the source detected with FAST, the second distribution peak time is smaller, suggesting that this peak is defined by the activity level of the source. We calculate the isotropic energy of the bursts using both a partial bandwidth and a full bandwidth and find that the energy distribution is not significantly changed. We find that an exponentially connected broken-power-law function can fit the cumulative burst energy distribution well, with the lower and higher-energy indices being $-1.22\pm0.01$ and $-4.27\pm0.23$, respectively. Assuming a radio radiative efficiency of $\eta_r = 10^{-4}$, the total isotropic energy of the bursts released during the four days when the source was active is already $3.9\times10^{46}$ erg, exceeding $\sim 23\%$ of the available magnetar dipolar magnetic energy. This challenges the magnetar models invoking an inefficient radio emission (e.g. synchrotron maser models)., Comment: 26 pages, 7 figures, accepted for publication in Research in Astronomy and Astrophysics

We report the properties of more than 800 bursts detected from the repeating fast radio burst (FRB) source FRB 20201124A with the Five-hundred-meter Aperture Spherical radio telescope (FAST) during an extremely active episode on UTC September 25th-28th, 2021 in a series of four papers. In this fourth paper of the series, we present a systematic search of the spin period and linear acceleration of the source object from both 996 individual pulse peaks and the dedispersed time series. No credible spin period was found from this data set. We rule out the presence of significant periodicity in the range between 1 ms to 100 s with a pulse duty cycle $< 0.49\pm0.08$ (when the profile is defined by a von-Mises function, not a boxcar function) and linear acceleration up to $300$ m s$^{-2}$ in each of the four one-hour observing sessions, and up to $0.6$ m s$^{-2}$ in all 4 days. These searches contest theoretical scenarios involving a 1 ms to 100 s isolated magnetar/pulsar with surface magnetic field $<10^{15}$ G and a small duty cycle (such as in a polar-cap emission mode) or a pulsar with a companion star or black hole up to 100 M$_{\rm \odot}$ and $P_b>10$ hours. We also perform a periodicity search of the fine structures and identify 53 unrelated millisecond-timescale "periods" in multi-components with the highest significance of 3.9 $\sigma$. The "periods" recovered from the fine structures are neither consistent nor harmonically related. Thus they are not likely to come from a spin period. We caution against claiming spin periodicity with significance below $\sim$ 4 $\sigma$ with multi-components from one-off FRBs. We discuss the implications of our results and the possible connections between FRB multi-components and pulsar micro-structures., Comment: Accepted by Research in Astronomy and Astrophysics (RAA)

As the third paper in the multiple-part series, we report the statistical properties of radio bursts detected from the repeating fast radio burst (FRB) source FRB 20201124A with the Five-hundred-meter Aperture Spherical radio telescope (FAST) during an extremely active episode between the 25th and the 28th of September 2021 (UT). We focus on the polarisation properties of 536 bright bursts with $\mathrm{S/N}>50$. We found that the Faraday rotation measures (RMs) monotonically dropped from $-579 \ {\rm rad \ m^{-2}}$ to $-605 \ {\rm rad \ m^{-2}}$ in the 4-day window. The RM values were compatible with the values ($-300$ to $-900\ {\rm rad \ m^{-2}}$ ) reported 4 month ago (Xu et al. 2022). However, the RM evolution rate in the current observation window was at least an order of magnitude smaller than the one ($\sim 500\ {\rm rad \ m^{-2}\, day^{-1}}$) previously reported during the rapid RM-variation phase, but is still higher than the one ($\le 1\ {\rm rad \ m^{-2} day^{-1}}$ ) during the later RM no-evolution phase. The bursts of FRB 20201124A were highly polarised with the total degree of polarisation (circular plus linear) greater than 90% for more than 90\% of all bursts. The distribution of linear polarisation position angles (PAs), degree of linear polarisation ($L/I$), and degree of circular polarisation ($V/I$) can be characterised with unimodal distribution functions. During the observation window, the distributions became wider with time, i.e. with larger scatter, but the centroids of the distribution functions remained nearly constant. For individual bursts, significant PA variations (confidence level 5-$\sigma$) were observed in 33% of all bursts. The polarisation of single pulses seems to follow certain complex trajectories on the Poincar\'e sphere, which may shed light on the radiation mechanism at the source or the plasma properties along the path of FRB propagation., Comment: 25 pages, 16 figures. Accepted by Research in Astronomy and Astrophysics (RAA)