arXiv subjects

Song Feng

Publications and source records attributed to Song Feng.

At least 37 records · Page 2Linked to original sources

arXiv2021

Fine-grained Solar Flare Forecasting Based on the Hybrid Convolutional Neural Networks

Improving the performance of solar flare forecasting is a hot topic in solar physics research field. Deep learning has been considered a promising approach to perform solar flare forecasting in recent years. We first used the Generative Adversarial Networks (GAN) technique augmenting sample data to balance samples with different flare classes. We then proposed a hybrid convolutional neural network (CNN) model M for forecasting flare eruption in a solar cycle. Based on this model, we further investigated the effects of the rising and declining phases for flare forecasting. Two CNN models, i.e., Mrp and Mdp, were presented to forecast solar flare eruptions in the rising phase and declining phase of solar cycle 24, respectively. A series of testing results proved: 1) Sample balance is critical for the stability of the CNN model. The augmented data generated by GAN effectively improved the stability of the forecast model. 2) For C-class, M-class, and X-class flare forecasting using Solar Dynamics Observatory (SDO) line-of-sight (LOS) magnetograms, the means of true skill statistics (TSS) score of M are 0.646, 0.653 and 0.762, which improved by 20.1%, 22.3%, 38.0% compared with previous studies. 3) It is valuable to separately model the flare forecasts in the rising and declining phases of a solar cycle. Compared with model M, the means of TSS score for No-flare, C-class, M-class, X-class flare forecasting of the Mrp improved by 5.9%, 9.4%, 17.9% and 13.1%, and the Mdp improved by 1.5%, 2.6%, 11.5% and 12.2%.

astro-ph.SR↗

arXiv2021

MultiDoc2Dial: Modeling Dialogues Grounded in Multiple Documents

We propose MultiDoc2Dial, a new task and dataset on modeling goal-oriented dialogues grounded in multiple documents. Most previous works treat document-grounded dialogue modeling as a machine reading comprehension task based on a single given document or passage. In this work, we aim to address more realistic scenarios where a goal-oriented information-seeking conversation involves multiple topics, and hence is grounded on different documents. To facilitate such a task, we introduce a new dataset that contains dialogues grounded in multiple documents from four different domains. We also explore modeling the dialogue-based and document-based context in the dataset. We present strong baseline approaches and various experimental results, aiming to support further research efforts on such a task.

cs.CL↗

arXiv2021

Dynamics and Sensitivity of Signaling Pathways

Signaling pathways serve to communicate information about extracellular conditions into the cell, to both the nucleus and cytoplasmic processes to control cell responses. Genetic mutations in signaling network components are frequently associated with cancer and can result in cells acquiring an ability to divide and grow uncontrollably. Because signaling pathways play such a significant role in cancer initiation and advancement, their constituent proteins are attractive therapeutic targets. In this review, we discuss how signaling pathway modeling can assist with identifying effective drugs for treating diseases, such as cancer. An achievement that would facilitate the use of such models is their ability to identify controlling biochemical parameters in signaling pathways, such as molecular abundances and chemical reaction rates, because this would help determine effective points of attack by therapeutics.

q-bio.MN↗

arXiv2021

MultiXNet: Multiclass Multistage Multimodal Motion Prediction

One of the critical pieces of the self-driving puzzle is understanding the surroundings of a self-driving vehicle (SDV) and predicting how these surroundings will change in the near future. To address this task we propose MultiXNet, an end-to-end approach for detection and motion prediction based directly on lidar sensor data. This approach builds on prior work by handling multiple classes of traffic actors, adding a jointly trained second-stage trajectory refinement step, and producing a multimodal probability distribution over future actor motion that includes both multiple discrete traffic behaviors and calibrated continuous position uncertainties. The method was evaluated on large-scale, real-world data collected by a fleet of SDVs in several cities, with the results indicating that it outperforms existing state-of-the-art approaches.

cs.CV↗

arXiv2021

Explaining Neural Network Predictions on Sentence Pairs via Learning Word-Group Masks

Explaining neural network models is important for increasing their trustworthiness in real-world applications. Most existing methods generate post-hoc explanations for neural network models by identifying individual feature attributions or detecting interactions between adjacent features. However, for models with text pairs as inputs (e.g., paraphrase identification), existing methods are not sufficient to capture feature interactions between two texts and their simple extension of computing all word-pair interactions between two texts is computationally inefficient. In this work, we propose the Group Mask (GMASK) method to implicitly detect word correlations by grouping correlated words from the input text pair together and measure their contribution to the corresponding NLP tasks as a whole. The proposed method is evaluated with two different model architectures (decomposable attention model and BERT) across four datasets, including natural language inference and paraphrase identification tasks. Experiments show the effectiveness of GMASK in providing faithful explanations to these models.

cs.CL↗

arXiv2020

doc2dial: A Goal-Oriented Document-Grounded Dialogue Dataset

We introduce doc2dial, a new dataset of goal-oriented dialogues that are grounded in the associated documents. Inspired by how the authors compose documents for guiding end users, we first construct dialogue flows based on the content elements that corresponds to higher-level relations across text sections as well as lower-level relations between discourse units within a section. Then we present these dialogue flows to crowd contributors to create conversational utterances. The dataset includes about 4800 annotated conversations with an average of 14 turns that are grounded in over 480 documents from four domains. Compared to the prior document-grounded dialogue datasets, this dataset covers a variety of dialogue scenes in information-seeking conversations. For evaluating the versatility of the dataset, we introduce multiple dialogue modeling tasks and present baseline approaches.

cs.CL↗

arXiv2020

Hypergraph Models of Biological Networks to Identify Genes Critical to Pathogenic Viral Response

Background: Representing biological networks as graphs is a powerful approach to reveal underlying patterns, signatures, and critical components from high-throughput biomolecular data. However, graphs do not natively capture the multi-way relationships present among genes and proteins in biological systems. Hypergraphs are generalizations of graphs that naturally model multi-way relationships and have shown promise in modeling systems such as protein complexes and metabolic reactions. In this paper we seek to understand how hypergraphs can more faithfully identify, and potentially predict, important genes based on complex relationships inferred from genomic expression data sets. Results: We compiled a novel data set of transcriptional host response to pathogenic viral infections and formulated relationships between genes as a hypergraph where hyperedges represent significantly perturbed genes, and vertices represent individual biological samples with specific experimental conditions. We find that hypergraph betweenness centrality is a superior method for identification of genes important to viral response when compared with graph centrality. Conclusions: Our results demonstrate the utility of using hypergraphs to represent complex biological systems and highlight central important responses in common to a variety of highly pathogenic viruses.

q-bio.QM↗

arXiv2020

Learning Lane Graph Representations for Motion Forecasting

We propose a motion forecasting model that exploits a novel structured map representation as well as actor-map interactions. Instead of encoding vectorized maps as raster images, we construct a lane graph from raw map data to explicitly preserve the map structure. To capture the complex topology and long range dependencies of the lane graph, we propose LaneGCN which extends graph convolutions with multiple adjacency matrices and along-lane dilation. To capture the complex interactions between actors and maps, we exploit a fusion network consisting of four types of interactions, actor-to-lane, lane-to-lane, lane-to-actor and actor-to-actor. Powered by LaneGCN and actor-map interactions, our model is able to predict accurate and realistic multi-modal trajectories. Our approach significantly outperforms the state-of-the-art on the large scale Argoverse motion forecasting benchmark.

cs.CV↗

arXiv2020

Preflare very long-periodic pulsations observed in Halpha emission before the onset of a solar flare

Very long-periodic pulsations during preflare phases (preflare-VLPs) have been detected in the full-disk solar soft X-Ray (SXR) flux (see Tan et al. 2016). They may be regarded as precursors to solar flares and may help us better understand the trigger mechanism of solar flares. In this letter, we report a preflare-VLP event before the onset of an M1.1 circular-ribbon flare on 2015 October 16. It was simultaneously observed in Halpha, SXR, and extreme ultraviolet (EUV) wavelengths, which were recorded by the NVST, GOES, EVE, and AIA respectively. The preflare-VLP is identified as the repeat and quasi-periodic pulses in light curves during preflare phase, which might be modulated by LRC-circuit oscillation in the current-carrying plasma loop. The quasi-periodicity can be determined from the Fourier power spectrum with Markov chain Monte Carlo (MCMC)-based Bayesian (e.g., Liang et al. 2020), such as ~9.3 minutes. We present the first report of a preflare-VLP event in the local Halpha line and EUV wavelength, which could be considered a precursor of a solar flare. This finding should therefore prove useful for the prediction of solar flares, especially for powerful flares.

astro-ph.SR↗

arXiv2020

Generation of Solar Spicules and Subsequent Atmospheric Heating

Spicules are rapidly evolving fine-scale jets of magnetized plasma in the solar chromosphere. It remains unclear how these prevalent jets originate from the solar surface and what role they play in heating the solar atmosphere. Using the Goode Solar Telescope at the Big Bear Solar Observatory, we observed spicules emerging within minutes of the appearance of opposite-polarity magnetic flux around dominant-polarity magnetic field concentrations. Data from the Solar Dynamics Observatory showed subsequent heating of the adjacent corona. The dynamic interaction of magnetic fields (likely due to magnetic reconnection) in the partially ionized lower solar atmosphere appears to generate these spicules and heat the upper solar atmosphere.

astro-ph.SR↗

arXiv2020

Magnetic connectivity between the light bridge and penumbra in a sunspot

A light bridge is a prominent structure commonly observed within a sunspot. Its presence usually triggers a wealth of dynamics in a sunspot, and has a lasting impact on sunspot evolution. However, the fundamental structure of light bridges is still not well understood. In this study, we used the high-resolution spectropolarimetry data obtained by the Solar Optical Telescope onboard the Hinode satellite to analyze the magnetic and thermal structure of a light bridge at \AR. We also combined the high-cadence $1700\unitÅ$ channel data provided by the Atmospheric Imaging Assembly onboard the Solar Dynamic Observatory to study the dynamics on this bridge. We found that a pair of blue and red Doppler shift patches at two ends of this bridge, this pattern appears to be the convective motion directed by the horizontal component of the magnetic field aligned with the spine of this bridge. Paired upward and downward motions implies that the light bridge could have a two-legged or undulate magnetic field. Significant four minute oscillations in the emission intensity of the $1700\unitÅ$ bandpass were detected at two ends, which had overlap with the paired blue and red shift patches. The oscillatory signals at the light bridge and the penumbra were highly correlated with each other. Although they are separated in space at the photosphere, the periodicity seems to have a common origin from the underneath. Therefore, we infer that the light bridge and penumbra could share a common magnetic source and become fragmented at the photosphere by magneto-convection.

astro-ph.SR↗

arXiv2020

Estimating red noise in quasi-periodic signals with MCMC-based Bayesian

Multi-parameter Bayesian inferences based on Markov chain Monte Carlo (MCMC) samples have been widely used to estimate red noise in solar period-periodic signals. To MCMC, proper priors and sufficient iterations are prerequisites ensuring the accuracy of red noise estimation. We used MCMC-based Bayesian inferences to estimate 100 groups of red noise synthesized randomly for evaluating its accuracy. At the same time, the Brooks-Gelman algorithm was employed to precisely diagnose the convergence of the Markov chains generated by MCMC. The root-mean-square error of parameter inferences to the synthetic data is only 1.14. Furthermore, we applied the algorithm to analyze the oscillation modes in a sunspot and a flare. A 70 s period is detected in the sunspot umbra in addition to 3- and 5-minute periods, and a 40 s period is detected in the flare. The results prove that estimating red noise with MCMC-based Bayesian has more high accuracy in the case of proper priors and convergence. We also find that the number of iterations increases dramatically to achieve convergence as the number of parameters grows. Therefore, we strongly recommend that when estimating red noise with MCMC-based Bayesian, different initial values must be selected to ensure that the entire posterior distribution is covered.

astro-ph.SR↗

arXiv2020

Propagating Slow Sausage Waves in a Sunspot Observed by the New Vacuum Solar Telescope

A sunspot is an ideal waveguide for a variety of magnetohydrodynamic waves, which carry a significant amount of energy to the upper atmosphere and could be used as a tool to probe magnetic and thermal structure of a sunspot. In this study, we used the New Vacuum Solar Telescope and took high-resolution image sequences simultaneously in both TiO (7058$\pm$10 Å) and H$_α$ (6562$\pm$2.5 Å) bandpasses. We extracted the area and total emission intensity variations of sunspot umbra and analyzed the signals with synchrosqueezing transform. We found that the area and emission intensity varied with both three and five minute periodicity. Moreover, the area and intensity oscillated in phase with each other, this fact hold in both TiO and H$_α$ data. We interpret this oscillatory signal as propagating slow sausage wave. The propagation speed is estimated at about 8 km$\cdot$s$^{-1}$. We infer that this sunspot's umbra could have temperature as low as 2800--3500 K.

astro-ph.SR↗

arXiv2019

A Compact Source for Quasi-Periodic Pulsation in an M-class Solar Flare

Quasi-periodic pulsations (QPP) are usually found in the light curves of solar and stellar flares, they carry the features of time characteristics and plasma emission of the flaring core, and could be used to diagnose the coronas of the Sun and remote stars. In this study, we combined the Atmospheric Imaging Assembly (AIA) onboard the Solar Dynamics Observatory and the Nobeyama Radioheliograph (NoRH) to observe an M7.7 class flare occurred at active region 11520 on 19 July 2012. A QPP was detected both in the AIA $131\unit{Å}$ bandpass and the NoRH $17\unit{GHz}$ channel, it had a period of about four minutes. In the spatial distribution of Fourier power, we found that this QPP originated from a compact source and that it overlapped with the X-ray source above the loop top. The plasma emission intensities in the AIA $131\unit{Å}$ bandpass were highly correlated within this region. The source region is further segmented into stripes that oscillated with distinctive phases. Evidence in this event suggests that this QPP was likely to be generated by intermittent energy injection into the reconnection region.

astro-ph.SR↗

arXiv2019

Drill-down: Interactive Retrieval of Complex Scenes using Natural Language Queries

This paper explores the task of interactive image retrieval using natural language queries, where a user progressively provides input queries to refine a set of retrieval results. Moreover, our work explores this problem in the context of complex image scenes containing multiple objects. We propose Drill-down, an effective framework for encoding multiple queries with an efficient compact state representation that significantly extends current methods for single-round image retrieval. We show that using multiple rounds of natural language queries as input can be surprisingly effective to find arbitrarily specific images of complex scenes. Furthermore, we find that existing image datasets with textual captions can provide a surprisingly effective form of weak supervision for this task. We compare our method with existing sequential encoding and embedding networks, demonstrating superior performance on two proposed benchmarks: automatic image retrieval on a simulated scenario that uses region captions as queries, and interactive image retrieval using real queries from human evaluators.

cs.CV↗

arXiv2019

Discrete Residual Flow for Probabilistic Pedestrian Behavior Prediction

Self-driving vehicles plan around both static and dynamic objects, applying predictive models of behavior to estimate future locations of the objects in the environment. However, future behavior is inherently uncertain, and models of motion that produce deterministic outputs are limited to short timescales. Particularly difficult is the prediction of human behavior. In this work, we propose the discrete residual flow network (DRF-Net), a convolutional neural network for human motion prediction that captures the uncertainty inherent in long-range motion forecasting. In particular, our learned network effectively captures multimodal posteriors over future human motion by predicting and updating a discretized distribution over spatial locations. We compare our model against several strong competitors and show that our model outperforms all baselines.

cs.CV↗

arXiv2019

Spatial Distributions of Sunspot Oscillation Modes at Different Temperatures

Three- and five-minute oscillations of sunspots have different spatial distributions in the solar atmospheric layers. The spatial distributions are crucial to reveal the physical origin of sunspot oscillations and to investigate their propagation. In this study, six sunspots observed by Solar Dynamics Observatory/Atmospheric Imaging Assembly were used to obtain the spatial distributions of three- and five-minute oscillations. The fast Fourier transform method is applied to represent the power spectra of oscillation modes. We find that, from the temperature minimum to the lower corona, the powers of the five-minute oscillation exhibit a circle-shape distribution around its umbra, and the shapes gradually expand with temperature increase. However, the circle-shape is disappeared and the powers of the oscillations appear to be very disordered in the higher corona. This indicates that the five-minute oscillation can be suppressed in the high-temperature region. For the three-minute oscillations, from the temperature minimum to the high corona, their powers mostly distribute within an umbra, and part of them locate at the coronal fan loop structures. Moreover, those relative higher powers are mostly concentrated in the position of coronal loop footpoints.

astro-ph.SR↗

arXiv2019

Text2Scene: Generating Compositional Scenes from Textual Descriptions

In this paper, we propose Text2Scene, a model that generates various forms of compositional scene representations from natural language descriptions. Unlike recent works, our method does NOT use Generative Adversarial Networks (GANs). Text2Scene instead learns to sequentially generate objects and their attributes (location, size, appearance, etc) at every time step by attending to different parts of the input text and the current status of the generated scene. We show that under minor modifications, the proposed framework can handle the generation of different forms of scene representations, including cartoon-like scenes, object layouts corresponding to real images, and synthetic images. Our method is not only competitive when compared with state-of-the-art GAN-based methods using automatic metrics and superior based on human judgments but also has the advantage of producing interpretable results.

cs.CV↗