arXiv subjects

Bo Zhao

Publications and source records attributed to Bo Zhao.

At least 307 records · Page 17Linked to original sources

arXiv2018

On the Effects of Self-Obscuration in the (Sub-)Millimeter Spectral Indices and Appearance of Protostellar Disks

In this paper we explore the effects of self-obscuration in protostellar disks with a radially decreasing temperature gradient and a colder midplane. We are motivated by recent reports of resolved dark lanes (`hamburgers') and (sub)mm spectral indices systematically below the ISM value for optically thin dust $α_{\rm ISM} =3.7$. We explore several model grids, scaling disk mass and varying inclination angle $i$ and observing frequency $ν$ from the VLA Ka band ($\sim 37$ GHz) to ALMA Band 8 ($\sim 405$ GHz). We also consider the effects of decreasing the index of the (sub-)mm dust opacity power law $β$ from 1.7 to 1. We find that a distribution of disk masses in the range $M_{\rm disk} = 0.01-2~M_\odot$ is needed to reproduce the observed distribution of spectral indices, and that assuming a fixed $β=1.7$ gives better results than $β=1$. A wide distribution of disk masses is also needed to produce some cases with $α<2$, as reported for some sources in the literature. Such extremely low spectral indices arise naturally when the selected observing frequencies sample the appropriate change in the temperature structure of the optically thick model disk. Our results show that protostellar disk masses could often be underestimated by $> \times10$, and are consistent with recent hydrodynamical simulations. Although we do not rule out the possibility of some grain growth occurring within the short protostellar timescales, we conclude that self-obscuration needs to be taken into account.

astro-ph.GA↗

arXiv2018

Observation of magnetically tunable Feshbach resonances in ultracold $^{23}$Na$^{40}$K+$^{40}$K collisions

Resonances in ultracold collisions involving heavy molecules are difficult to understand, and have proven challenging to detect. Here we report the observation of magnetically tunable Feshbach resonances in ultracold collisions between $^{23}$Na$^{40}$K molecules in the rovibrational ground state and $^{40}$K atoms. We prepare the atoms and molecules in various hyperfine levels of their ground states and observe the loss of molecules as a function of the magnetic field. The atom-molecule Feshbach resonances are identified by observing an enhancement of the loss rate coefficients. We have observed three resonances at approximately 101 G in various atom-molecule scattering channels, with the widths being a few hundred milliGauss. The observed atom-molecule Feshbach resonances at ultralow temperatures probe the three-body potential energy surface with an unprecedented resolution. Our work will help to improve the understanding of complicated ultracold collisions, and open up the possibility of creating ultracold triatomic molecules.

physics.atom-ph↗

arXiv2018

MSplit LBI: Realizing Feature Selection and Dense Estimation Simultaneously in Few-shot and Zero-shot Learning

It is one typical and general topic of learning a good embedding model to efficiently learn the representation coefficients between two spaces/subspaces. To solve this task, $L_{1}$ regularization is widely used for the pursuit of feature selection and avoiding overfitting, and yet the sparse estimation of features in $L_{1}$ regularization may cause the underfitting of training data. $L_{2}$ regularization is also frequently used, but it is a biased estimator. In this paper, we propose the idea that the features consist of three orthogonal parts, \emph{namely} sparse strong signals, dense weak signals and random noise, in which both strong and weak signals contribute to the fitting of data. To facilitate such novel decomposition, \emph{MSplit} LBI is for the first time proposed to realize feature selection and dense estimation simultaneously. We provide theoretical and simulational verification that our method exceeds $L_{1}$ and $L_{2}$ regularization, and extensive experimental results show that our method achieves state-of-the-art performance in the few-shot and zero-shot learning.

cs.CV↗

arXiv2018

A Large-scale Attribute Dataset for Zero-shot Learning

Zero-Shot Learning (ZSL) has attracted huge research attention over the past few years; it aims to learn the new concepts that have never been seen before. In classical ZSL algorithms, attributes are introduced as the intermediate semantic representation to realize the knowledge transfer from seen classes to unseen classes. Previous ZSL algorithms are tested on several benchmark datasets annotated with attributes. However, these datasets are defective in terms of the image distribution and attribute diversity. In addition, we argue that the "co-occurrence bias problem" of existing datasets, which is caused by the biased co-occurrence of objects, significantly hinders models from correctly learning the concept. To overcome these problems, we propose a Large-scale Attribute Dataset (LAD). Our dataset has 78,017 images of 5 super-classes, 230 classes. The image number of LAD is larger than the sum of the four most popular attribute datasets. 359 attributes of visual, semantic and subjective properties are defined and annotated in instance-level. We analyze our dataset by conducting both supervised learning and zero-shot learning tasks. Seven state-of-the-art ZSL algorithms are tested on this new dataset. The experimental results reveal the challenge of implementing zero-shot learning on our dataset.

cs.CV↗

arXiv2018

Modular Generative Adversarial Networks

Existing methods for multi-domain image-to-image translation (or generation) attempt to directly map an input image (or a random vector) to an image in one of the output domains. However, most existing methods have limited scalability and robustness, since they require building independent models for each pair of domains in question. This leads to two significant shortcomings: (1) the need to train exponential number of pairwise models, and (2) the inability to leverage data from other domains when training a particular pairwise mapping. Inspired by recent work on module networks, this paper proposes ModularGAN for multi-domain image generation and image-to-image translation. ModularGAN consists of several reusable and composable modules that carry on different functions (e.g., encoding, decoding, transformations). These modules can be trained simultaneously, leveraging data from all domains, and then combined to construct specific GAN networks at test time, according to the specific image translation task. This leads to ModularGAN's superior flexibility of generating (or translating to) an image in any desired domain. Experimental results demonstrate that our model not only presents compelling perceptual results but also outperforms state-of-the-art methods on multi-domain facial attribute transfer.

cs.CV↗

arXiv2018

Left-Right Comparative Recurrent Model for Stereo Matching

Leveraging the disparity information from both left and right views is crucial for stereo disparity estimation. Left-right consistency check is an effective way to enhance the disparity estimation by referring to the information from the opposite view. However, the conventional left-right consistency check is an isolated post-processing step and heavily hand-crafted. This paper proposes a novel left-right comparative recurrent model to perform left-right consistency checking jointly with disparity estimation. At each recurrent step, the model produces disparity results for both views, and then performs online left-right comparison to identify the mismatched regions which may probably contain erroneously labeled pixels. A soft attention mechanism is introduced, which employs the learned error maps for better guiding the model to selectively focus on refining the unreliable regions at the next recurrent step. In this way, the generated disparity maps are progressively improved by the proposed recurrent model. Extensive evaluations on KITTI 2015, Scene Flow and Middlebury benchmarks validate the effectiveness of our model, demonstrating that state-of-the-art stereo disparity estimation results can be achieved by this new model.

cs.CV↗

arXiv2018

Constraint on ion-neutral drift velocity in the Class 0 protostar B335 from ALMA observations

Ambipolar diffusion can cause a velocity drift between ions and neutrals. This is one of the non-ideal MHD effects proposed to enable the formation of Keplerian disks with sizes of tens of au. To observationally study ambipolar diffusion in collapsing protostellar envelopes, we compare gas kinematics traced by the H13CO+ (3-2) and C18O (2-1) lines in the Class 0 protostar B335 obtained with our ALMA observations. A central compact (~1"-2") component that is elongated perpendicular to the outflow direction and exhibits a clear velocity gradient along the outflow direction is observed in both lines and most likely traces the infalling flattened envelope. We constructed kinematical models to fit the observed velocity structures and to measure the infalling velocities of the ionized and neutral gas on a 100 au scale in B335. The infalling velocities in the H13CO+ and C18O emission are both measured to be 0.85+/-0.2 km/s at a radius of 100 au, suggesting that the velocity drift between the ionized and neutral gas is at most 0.3 km/s at a radius of 100 au. The Hall parameter for H13CO+ is estimated to be >>1 on a 100 au scale in B335, so that H13CO+ is expected to be attached to the magnetic field. Our non-detection or upper limit of the velocity drift between ions and neutrals could suggest that the magnetic field remains rather well coupled to the bulk neutral material on a 100 au scale in this source, and that any significant field-matter decoupling, if present, likely occurs only on a smaller scale, leading to an accumulation of magnetic flux and thus efficient magnetic braking in the inner envelope. However, because of our limited angular resolution, we cannot rule out a significant ambipolar drift only in the midplane of the infalling envelope. Future observations with higher angular resolutions (~0.1") are needed to examine this possibility and ambipolar diffusion on a smaller scale.

astro-ph.SR↗

arXiv2018

Effect of Grain Size on Differential Desorption of Volatile Species and on Non-ideal MHD Diffusivity

We developed a chemical network for modeling the chemistry and non-ideal MHD effects from the collapsing dense molecular clouds to protostellar disks. First, we re-formulated the cosmic-ray desorption rate by considering the variations of desorption rate over the grain size distribution. We find that the differential desorption of volatile species is amplified by the grains larger than 0.1 $μ$m, because larger grains are heated to a lower temperature by cosmic-rays and hence more sensitive to the variations in binding energies. As a result, atomic nitrogen N is $\sim$2 orders of magnitude more abundant than CO; N$_2$H$^+$ also becomes a few times more abundant than HCO$^+$ due to the increased gas-phase N$_2$. However, the changes in ionization fraction due to freeze-out and desorption only have minor effects on the non-ideal MHD diffusivities. Our chemical network confirms that the very small grains (VSGs: below a few 100 $Å$) weakens the efficiency of both ambipolar diffusion and Hall effect. In collapsing dense cores, a maximum ambipolar diffusion is achieved when truncating the MRN size distribution at 0.1 $μ$m, and for a maximum Hall effect, the truncation occurs at 0.04 $μ$m. We conclude that the grain size distribution is crucial to the differential depletion between CO and N$_2$ related molecules, as well as to the non-ideal MHD diffusivities in dense cores.

astro-ph.SR↗

arXiv2018

Multi-View Image Generation from a Single-View

This paper addresses a challenging problem -- how to generate multi-view cloth images from only a single view input. To generate realistic-looking images with different views from the input, we propose a new image generation model termed VariGANs that combines the strengths of the variational inference and the Generative Adversarial Networks (GANs). Our proposed VariGANs model generates the target image in a coarse-to-fine manner instead of a single pass which suffers from severe artifacts. It first performs variational inference to model global appearance of the object (e.g., shape and color) and produce a coarse image with a different view. Conditioned on the generated low resolution images, it then proceeds to perform adversarial learning to fill details and generate images of consistent details with the input. Extensive experiments conducted on two clothing datasets, MVC and DeepFashion, have demonstrated that images of a novel view generated by our model are more plausible than those generated by existing approaches, in terms of more consistent global appearance as well as richer and sharper details.

cs.CV↗

arXiv2017

Zero-shot Learning via Shared-Reconstruction-Graph Pursuit

Zero-shot learning (ZSL) aims to recognize objects from novel unseen classes without any training data. Recently, structure-transfer based methods are proposed to implement ZSL by transferring structural knowledge from the semantic embedding space to image feature space to classify testing images. However, we observe that such a knowledge transfer framework may suffer from the problem of the geometric inconsistency between the data in the training and testing spaces. We call this problem as the space shift problem. In this paper, we propose a novel graph based method to alleviate this space shift problem. Specifically, a Shared Reconstruction Graph (SRG) is pursued to capture the common structure of data in the two spaces. With the learned SRG, each unseen class prototype (cluster center) in the image feature space can be synthesized by the linear combination of other class prototypes, so that testing instances can be classified based on the distance to these synthesized prototypes. The SRG bridges the image feature space and semantic embedding space. By applying spectral clustering on the learned SRG, many meaningful clusters can be discovered, which interprets ZSL performance on the datasets. Our method can be easily extended to the generalized zero-shot learning setting. Experiments on three popular datasets show that our method outperforms other methods on all datasets. Even with a small number of training samples, our method can achieve the state-of-the-art performance.

cs.CV↗

arXiv2017

AI Challenger : A Large-scale Dataset for Going Deeper in Image Understanding

Significant progress has been achieved in Computer Vision by leveraging large-scale image datasets. However, large-scale datasets for complex Computer Vision tasks beyond classification are still limited. This paper proposed a large-scale dataset named AIC (AI Challenger) with three sub-datasets, human keypoint detection (HKD), large-scale attribute dataset (LAD) and image Chinese captioning (ICC). In this dataset, we annotate class labels (LAD), keypoint coordinate (HKD), bounding box (HKD and LAD), attribute (LAD) and caption (ICC). These rich annotations bridge the semantic gap between low-level images and high-level concepts. The proposed dataset is an effective benchmark to evaluate and improve different computational methods. In addition, for related tasks, others can also use our dataset as a new resource to pre-train their models.

cs.CV↗

arXiv2017

Decoupling of Magnetic Fields in Collapsing Protostellar Envelopes and Disk Formation and Fragmentation

Efficient magnetic braking is a formidable obstacle to the formation of rotationally supported disks (RSDs) around protostars in magnetized dense cores. We have previously shown, through 2D (axisymmetric) non-ideal MHD simulations, that removing very small grains (VSGs: ~10 Å$~$to few 100 Å) can greatly enhance ambipolar diffusion and enable the formation of RSDs. Here we extend the simulations of disk formation enabled by VSG removal to 3D. We find that the key to this scenario of disk formation is that the drift velocity of the magnetic field almost cancels out the infall velocity of the neutrals in the $10^2$-$10^3$ AU-scale "pseudo-disk" where the field lines are most severely pinched and most of protostellar envelope mass infall occurs. As a result, the bulk neutral envelope matter can collapse without dragging much magnetic flux into the disk-forming region, which lowers the magnetic braking efficiency. We find that the initial disks enabled by VSG removal tend to be Toomre-unstable, which leads to the formation of prominent spiral structures that function as centrifugal barriers. The piling-up of infall material near the centrifugal barrier often produces dense fragments of tens of Jupiter masses, especially in cores that are not too strongly magnetized. Some fragments accrete onto the central stellar object, producing bursts in mass accretion rate. Others are longer lived, although whether they can survive long-term to produce multiple systems remains to be ascertained. Our results highlight the importance of dust grain evolution in determining the formation and properties of protostellar disks and potentially multiple systems.

astro-ph.SR↗

arXiv2017

Feshbach Loss Spectroscopy in an Ultracold $^{23}$Na and $^{40}$K Mixture

We perform Feshbach spectroscopy in an ultracold mixture of $^{23}$Na and $^{40}$K with different spin-state combinations. We have observed 24 new interspecies Feshbach resonances at magnetic field up to 350 G. A full coupled-channel calculation is performed to assign these resonances. Among them, 12 resonances are identified as d-wave Feshbach resonances. These d-wave Feshbach resonances are about 5 G systematically smaller than the predications based on previous model potential. Taking into account these new experimental results, we improve the Born-Oppenheimer potentials between Na and K, and achieve good agreement between the theory and experiment for all the observed Feshbach resonances.

cond-mat.quant-gas↗

arXiv2017

State-to-state endothermic and nearly thermoneutral reactions in an ultracold atom-dimer mixture

Chemical reactions at ultracold temperature provide an ideal platform to study chemical reactivity at the fundamental level, and to understand how chemical reactions are governed by quantum mechanics. Recent years have witnessed the remarkable progress in studying ultracold chemistry with ultracold molecules. However, these works were limited to exothermic reactions. The direct observation of state-to-state ultracold endothermic reaction remains elusive. Here we report on the investigation of endothermic and nearly thermoneutral atom-exchange reactions in an ultracold atom-dimer mixture. By developing an indirect reactant-preparation method based on a molecular bound-bound transition, we are able to directly observe a universal endothermic reaction with tunable energy threshold and study the state-to-state reaction dynamics. The reaction rate coefficients show a strikingly threshold phenomenon. The influence of the reverse reaction on the reaction dynamics is observed for the endothermic and nearly thermoneutral reactions. We carry out zero-range quantum mechanical scattering calculations to obtain the reaction rate coefficients, and the three-body parameter is determined by comparison with the experiments. The observed endothermic and nearly thermoneutral reaction may be employed to implement collisional Sisyphus cooling of molecules, study the chemical reactions in degenerate quantum gases and conduct quantum simulation of Kondo effect with ultracold atoms.

cond-mat.quant-gas↗

arXiv2017

Neural Person Search Machines

We investigate the problem of person search in the wild in this work. Instead of comparing the query against all candidate regions generated in a query-blind manner, we propose to recursively shrink the search area from the whole image till achieving precise localization of the target person, by fully exploiting information from the query and contextual cues in every recursive search step. We develop the Neural Person Search Machines (NPSM) to implement such recursive localization for person search. Benefiting from its neural search mechanism, NPSM is able to selectively shrink its focus from a loose region to a tighter one containing the target automatically. In this process, NPSM employs an internal primitive memory component to memorize the query representation which modulates the attention and augments its robustness to other distracting regions. Evaluations on two benchmark datasets, CUHK-SYSU Person Search dataset and PRW dataset, have demonstrated that our method can outperform current state-of-the-arts in both mAP and top-1 evaluation protocols.

cs.CV↗

arXiv2017

Molecular outflow launched beyond the disk edge

One of the long-standing problems of star formation is the excess of angular momentum of the parent molecular cloud. In the classical picture, a fraction of angular momentum of the circumstellar material is removed by the magneto-centrifugally driven disk wind that is launched from a wide region throughout the disk. In this work, we investigate the kinematics in the envelope-disk transition zone of the Class I object BHB07-11, in the B59 core. For this purpose, we used the Atacama Large Millimeter/submillimeter Array in extended configuration to observe the thermal dust continuum emission ($λ_0 \sim$ 1.3 mm) and molecular lines (CO, C$^{18}$O and H$_2$CO), which are suitable tracers of disk, envelope, and outflow dynamics at a spatial resolution of $\sim 30$ AU. We report a bipolar outflow that was launched at symmetric positions with respect to the disk ($\sim$80~AU in radius), but was concentrated at a distance of 90--130~AU from the disk center. The two outflow lobes had a conical shape and the gas inside was accelerating. The large offset of the launching position coincided with the landing site of the infall material from the extended spiral structure (seen in dust) onto the disk. This indicates that bipolar outflows are efficiently launched within a narrow region outside the disk edge. We also identify a sharp transition in the gas kinematics across the tip of the spiral structure, which pinpoints the location of the so-called centrifugal barrier.

astro-ph.SR↗

arXiv2017

Phase transitions and spin excitations of spin-1 bosons in optical lattice

We investigate ground state properties of spin-1 bosonic system trapped in optical lattice with extended standard basis operator (SBO) method. For both ferromagnetic ($U_2<0$) and antiferromagnetic ($U_2>0$) systems, we analytically figure out the symmetry properties in Mott-insulator and superfluid phases, which would provide a deeper insight into the MI-SF phase transition process. Then by applying self-consistent approach to the method, we include the effect of quantum and thermal fluctuations and derive the MI-SF transition phase diagram, which is in quantitative agreement with recent Monte-Carlo simulation at zero temperature, and at finite temperature, we find the underestimation of finite-temperature-effect in the mean-field approximation method. If we further consider the spin excitations in the insulating states of spin-1 system in external field, distinct spin phases are expected. Therefore, in the Mott lobes with $n=1$ and $n=2$ atoms per site, we give analytical and numerical boundaries of the singlet, nematic, partially magnetic and ferromagnetic phases in the magnetic phase diagrams.

quant-ph↗

arXiv2017

Diversified Visual Attention Networks for Fine-Grained Object Classification

Fine-grained object classification is a challenging task due to the subtle inter-class difference and large intra-class variation. Recently, visual attention models have been applied to automatically localize the discriminative regions of an image for better capturing critical difference and demonstrated promising performance. However, without consideration of the diversity in attention process, most of existing attention models perform poorly in classifying fine-grained objects. In this paper, we propose a diversified visual attention network (DVAN) to address the problems of fine-grained object classification, which substan- tially relieves the dependency on strongly-supervised information for learning to localize discriminative regions compared with attentionless models. More importantly, DVAN explicitly pursues the diversity of attention and is able to gather discriminative information to the maximal extent. Multiple attention canvases are generated to extract convolutional features for attention. An LSTM recurrent unit is employed to learn the attentiveness and discrimination of attention canvases. The proposed DVAN has the ability to attend the object from coarse to fine granularity, and a dynamic internal representation for classification is built up by incrementally combining the information from different locations and scales of the image. Extensive experiments con- ducted on CUB-2011, Stanford Dogs and Stanford Cars datasets have demonstrated that the proposed diversified visual attention networks achieve competitive performance compared to the state- of-the-art approaches, without using any prior knowledge, user interaction or external resource in training or testing.

cs.CV↗