arXiv subjects

Bo Zhang

Publications and source records attributed to Bo Zhang.

At least 703 records · Page 39Linked to original sources

arXiv2019

Regularized Adversarial Sampling and Deep Time-aware Attention for Click-Through Rate Prediction

Improving the performance of click-through rate (CTR) prediction remains one of the core tasks in online advertising systems. With the rise of deep learning, CTR prediction models with deep networks remarkably enhance model capacities. In deep CTR models, exploiting users' historical data is essential for learning users' behaviors and interests. As existing CTR prediction works neglect the importance of the temporal signals when embed users' historical clicking records, we propose a time-aware attention model which explicitly uses absolute temporal signals for expressing the users' periodic behaviors and relative temporal signals for expressing the temporal relation between items. Besides, we propose a regularized adversarial sampling strategy for negative sampling which eases the classification imbalance of CTR data and can make use of the strong guidance provided by the observed negative CTR samples. The adversarial sampling strategy significantly improves the training efficiency, and can be co-trained with the time-aware attention model seamlessly. Experiments are conducted on real-world CTR datasets from both in-station and out-station advertising places.

cs.LG↗

arXiv2019

Exploring the spectral \textit{information content} in the LAMOST medium-resolution survey (MRS)

Low-resolution spectra are proved competitive to high-resolution spectra in determining many stellar labels at comparable precision. It is useful to consider the spectral information content when assessing the capability of a stellar spectrum in deriving precise stellar labels. In this work, we quantify the information content brought by the LAMOST-II medium-resolution spectroscopic survey (MRS) using the gradient spectra and the coefficients-of-dependence (CODs). In general, the wavelength coverage of the MRS well constrains the stellar labels but the sensitivities of different stellar labels vary with spectral types and metallicity of the stars of interest and, therefore, affect the performance of the stellar label determination from the MRS spectra. Applying the SLAM to the synthetic spectra which mimic the MRS data, we find the precision of the fundamental stellar parameters Teff, logg and [M/H] are better when combining both the blue and red bands of the MRS. This is especially important for warm stars since the H$α$ line located in the red part plays a more important role in determining the effective temperature for warm stars. With blue and red parts together, we are able to reach similar performance to the low-resolution spectra except for warm stars. However, at [M/H]$\sim-2.0$ dex, the uncertainties of fundamental stellar labels estimated from MRS are substantially larger than those from low-resolution spectra. We also tested the uncertainties of Teff, logg and [M/H] of from MRS data induced from the radial velocity mismatch and find that a mismatch of about 1 km s$^{-1}$, which is typical for LAMOST MRS data, would not significantly affect the stellar label estimates. At last, reference precision limits are calculated using synthetic gradient spectra, according to which we expect abundances of at least 17 elements to be measured precisely from MRS spectra.

astro-ph.SR↗

arXiv2019

Intrinsic magnetic topological insulator phases in the Sb doped MnBi2Te4 bulks and thin flakes

Magnetic topological insulators (MTIs) offer a combination of topologically nontrivial characteristics and magnetic order and show promise in terms of potentially interesting physical phenomena such as the quantum anomalous Hall (QAH) effect and topological axion insulating states. However, the understanding of their properties and potential applications have been limited due to a lack of suitable candidates for MTIs. Here, we grow two-dimensional single crystals of Mn(SbxBi(1-x))2Te4 bulk and exfoliate them into thin flakes in order to search for intrinsic MTIs. We perform angle-resolved photoemission spectroscopy, low-temperature transport measurements, and first-principles calculations to investigate the band structure, transport properties, and magnetism of this family of materials, as well as the evolution of their topological properties. We find that there exists an optimized MTI zone in the Mn(SbxBi(1-x))2Te4 phase diagram, which could possibly host a high-temperature QAH phase, offering a promising avenue for new device applications.

cond-mat.mtrl-sci↗

arXiv2019

Deriving the stellar labels of LAMOST spectra with Stellar LAbel Machine (SLAM)

The LAMOST survey has provided 9 million spectra in its Data Release 5 (DR5) at R$\sim$1800. Extracting precise stellar labels is crucial for such a large sample. In this paper, we report the implementation of the Stellar LAbel Machine (SLAM), which is a data-driven method based on Support Vector Regression (SVR), a robust non-linear regression technique. Thanks to the capability to model highly non-linear problems with SVR, SLAM generally can derive stellar labels over a wide range of spectral types. This gives it a unique capability compared to other popular data-driven methods. To illustrate this capability, we test the performance of SLAM on stars ranging from Teff$\sim$4000 to $\sim$8000 K trained on LAMOST spectra and stellar labels. At g-band signal-to-noise ratio (SNRg) higher than 100, the random uncertainties of Teff, logg and [Fe/H] are 50 K, 0.09 dex, and 0.07 dex, respectively. We then set up another SLAM model trained by APOGEE and LAMOST common stars to demonstrate its capability of dealing with high dimensional problems. The spectra are from LAMOST DR5 and the stellar labels of the training set are from APOGEE DR15, including Teff, logg, [M/H],[$α$/M], [C/M], and [N/M]. The cross-validated scatters at SNRg$\sim$100 are 49 K, 0.10 dex, 0.037 dex,0.026 dex, 0.058 dex, and 0.106 dex for these parameters, respectively. This performance is at the same level as other up-to-date data-driven models. As a byproduct, we also provide the latest catalog of $\sim$1 million LAMOST DR5 K giant stars with SLAM-predicted stellar labels in this work.

astro-ph.SR↗

arXiv2019

Pruning from Scratch

Network pruning is an important research field aiming at reducing computational costs of neural networks. Conventional approaches follow a fixed paradigm which first trains a large and redundant network, and then determines which units (e.g., channels) are less important and thus can be removed. In this work, we find that pre-training an over-parameterized model is not necessary for obtaining the target pruned structure. In fact, a fully-trained over-parameterized model will reduce the search space for the pruned structure. We empirically show that more diverse pruned structures can be directly pruned from randomly initialized weights, including potential models with better performance. Therefore, we propose a novel network pruning pipeline which allows pruning from scratch. In the experiments for compressing classification models on CIFAR10 and ImageNet datasets, our approach not only greatly reduces the pre-training burden of traditional pruning methods, but also achieves similar or even higher accuracy under the same computation budgets. Our results facilitate the community to rethink the effectiveness of existing techniques used for network pruning.

cs.CV↗

arXiv2019

Tracing Kinematic and Chemical Properties of Sagittarius Stream by K-Giants, M-Giants, and BHB stars

We characterize the kinematic and chemical properties of $\sim$3,000 Sagittarius (Sgr) stream stars, including K-giants, M-giants, and BHBs, select from SEGUE-2, LAMOST, and SDSS separately in Integrals-of-Motion space. The orbit of Sgr stream is quite clear from the velocity vector in $X$-$Z$ plane. Stars traced by K-giants and M-giants present the apogalacticon of trailing steam is $\sim$ 100 kpc. The metallicity distributions of Sgr K-, M-giants, and BHBs present that the M-giants are on average the most metal-rich population, followed by K-giants and BHBs. All of the K-, M-giants, and BHBs indicate that the trailing arm is on average more metal-rich than leading arm, and the K-giants show that the Sgr debris is the most metal-poor part. The $α$-abundance of Sgr stars exhibits a similar trend with the Galactic halo stars at lower metallicity ([Fe/H] $<\sim$ $-$1.0 dex), and then evolve down to lower [$α$/Fe] than disk stars at higher metallicity, which is close to the evolution pattern of $α$-element of Milky Way dwarf galaxies. We find $V_Y$ and metallicity of K-giants have gradients along the direction of line-of-sight from the Galactic center in $X$-$Z$ plane, and the K-giants show that $V_Y$ increases with metallicity at [Fe/H] $>\sim-$1.5 dex. After dividing the Sgr stream into bright and faint stream according to their locations in equatorial coordinate, the K-giants and BHBs show that the bright and faint stream present different $V_Y$ and metallicities, the bright stream is on average higher in $V_Y$ and metallicity than the faint stream.

astro-ph.GA↗

arXiv2019

Curriculum Learning for Deep Generative Models with Clustering

Training generative models like Generative Adversarial Network (GAN) is challenging for noisy data. A novel curriculum learning algorithm pertaining to clustering is proposed to address this issue in this paper. The curriculum construction is based on the centrality of underlying clusters in data points. The data points of high centrality takes priority of being fed into generative models during training. To make our algorithm scalable to large-scale data, the active set is devised, in the sense that every round of training proceeds only on an active subset containing a small fraction of already trained data and the incremental data of lower centrality. Moreover, the geometric analysis is presented to interpret the necessity of cluster curriculum for generative models. The experiments on cat and human-face data validate that our algorithm is able to learn the optimal generative models (e.g. ProGAN) with respect to specified quality metrics for noisy data. An interesting finding is that the optimal cluster curriculum is closely related to the critical point of the geometric percolation process formulated in the paper.

cs.LG↗

arXiv2019

Document Rectification and Illumination Correction using a Patch-based CNN

We propose a novel learning method to rectify document images with various distortion types from a single input image. As opposed to previous learning-based methods, our approach seeks to first learn the distortion flow on input image patches rather than the entire image. We then present a robust technique to stitch the patch results into the rectified document by processing in the gradient domain. Furthermore, we propose a second network to correct the uneven illumination, further improving the readability and OCR accuracy. Due to the less complex distortion present on the smaller image patches, our patch-based approach followed by stitching and illumination correction can significantly improve the overall accuracy in both the synthetic and real datasets.

cs.CV↗

arXiv2019

A Data-Center FPGA Acceleration Platform for Convolutional Neural Networks

Intensive computation is entering data centers with multiple workloads of deep learning. To balance the compute efficiency, performance, and total cost of ownership (TCO), the use of a field-programmable gate array (FPGA) with reconfigurable logic provides an acceptable acceleration capacity and is compatible with diverse computation-sensitive tasks in the cloud. In this paper, we develop an FPGA acceleration platform that leverages a unified framework architecture for general-purpose convolutional neural network (CNN) inference acceleration at a data center. To overcome the computation bound, 4,096 DSPs are assembled and shaped as supertile units (SUs) for different types of convolution, which provide up to 4.2 TOP/s 16-bit fixed-point performance at 500 MHz. The interleaved-task-dispatching method is proposed to map the computation across the SUs, and the memory bound is solved by a dispatching-assembling buffering model and broadcast caches. For various non-convolution operators, a filter processing unit is designed for general-purpose filter-like/pointwise operators. In the experiment, the performances of CNN models running on server-class CPUs, a GPU, and an FPGA are compared. The results show that our design achieves the best FPGA peak performance and a throughput at the same level as that of the state-of-the-art GPU in data centers, with more than 50 times lower latency.

cs.CV↗

arXiv2019

Blind Geometric Distortion Correction on Images Through Deep Learning

We propose the first general framework to automatically correct different types of geometric distortion in a single input image. Our proposed method employs convolutional neural networks (CNNs) trained by using a large synthetic distortion dataset to predict the displacement field between distorted images and corrected images. A model fitting method uses the CNN output to estimate the distortion parameters, achieving a more accurate prediction. The final corrected image is generated based on the predicted flow using an efficient, high-quality resampling method. Experimental results demonstrate that our algorithm outperforms traditional correction methods, and allows for interesting applications such as distortion transfer, distortion exaggeration, and co-occurring distortion correction.

cs.CV↗

arXiv2019

On Recovery of a Bounded Elastic Body by Electromagnetic Far-Field Measurements

This paper is concerned with the problem of scattering of a time-harmonic electromagnetic field by a three-dimensional elastic body. General transmission conditions are considered to model the interaction between the electromagnetic field and the elastic body on the interface by assuming Voigt's model. The existence of a unique solution of the interaction problem is proved in an appropriate Sobolev space by employing a variational method together with the classical Fredholm alternative. The inverse problem is then considered, which is to recover the elastic body by the scattered wave-field. It is shown that the shape and location of the elastic body can be uniquely determined by the fixed energy magnetic (or electric) far-field measurements corresponding to incident plane waves with all polarizations.

math.AP↗

arXiv2019

Multi-Task Deep Learning with Dynamic Programming for Embryo Early Development Stage Classification from Time-Lapse Videos

Time-lapse is a technology used to record the development of embryos during in-vitro fertilization (IVF). Accurate classification of embryo early development stages can provide embryologists valuable information for assessing the embryo quality, and hence is critical to the success of IVF. This paper proposes a multi-task deep learning with dynamic programming (MTDL-DP) approach for this purpose. It first uses MTDL to pre-classify each frame in the time-lapse video to an embryo development stage, and then DP to optimize the stage sequence so that the stage number is monotonically non-decreasing, which usually holds in practice. Different MTDL frameworks, e.g., one-to-many, many-to-one, and many-to-many, are investigated. It is shown that the one-to-many MTDL framework achieved the best compromise between performance and computational cost. To our knowledge, this is the first study that applies MTDL to embryo early development stage classification from time-lapse videos.

eess.IV↗

arXiv2019

Convergence of the perfectly matched layer method for transient acoustic-elastic interaction above an unbounded rough surface

This paper is concerned with the time-dependent acoustic-elastic interaction problem associated with a bounded elastic body immersed in a homogeneous air or fluid above an unbounded rough surface. The well-posedness and stability of the problem are first established by using the Laplace transform and the energy method. A perfectly matched layer (PML) is then introduced to truncate the interaction problem above a finite layer containing the elastic body, leading to a PML problem in a finite strip domain. We further establish the existence, uniqueness and stability estimate of solutions to the PML problem. Finally, we prove the exponential convergence of the PML problem in terms of the thickness and parameter of the PML layer, based on establishing an error estimate between the DtN operators of the original problem and the PML problem.

math.AP↗

arXiv2019

Deep Exemplar-based Video Colorization

This paper presents the first end-to-end network for exemplar-based video colorization. The main challenge is to achieve temporal consistency while remaining faithful to the reference style. To address this issue, we introduce a recurrent framework that unifies the semantic correspondence and color propagation steps. Both steps allow a provided reference image to guide the colorization of every frame, thus reducing accumulated propagation errors. Video frames are colorized in sequence based on the colorization history, and its coherency is further enforced by the temporal consistency loss. All of these components, learned end-to-end, help produce realistic videos with good temporal stability. Experiments show our result is superior to the state-of-the-art methods both quantitatively and qualitatively.

cs.CV↗

arXiv2019

Uniqueness in inverse electromagnetic scattering problem with phaseless far-field data at a fixed frequency

This paper is concerned with uniqueness in inverse electromagnetic scattering with phaseless far-field pattern at a fixed frequency. In our previous work [{\em SIAM J. Appl. Math.} {\bf 78} (2018), 3024-3039], by adding a known reference ball into the acoustic scattering system, it was proved that the impenetrable obstacle and the index of refraction of an inhomogeneous medium can be uniquely determined by the acoustic phaseless far-field patterns generated by infinitely many sets of superpositions of two plane waves with different directions at a fixed frequency. In this paper, we extend these uniqueness results to the inverse electromagnetic scattering case. The phaseless far-field data are the modulus of the tangential component in the orientations $\mathbf{e}_ϕ$ and $\mathbf{e}_θ$, respectively, of the electric far-field pattern measured on the unit sphere and generated by infinitely many sets of superpositions of two electromagnetic plane waves with different directions and polarizations. Our proof is mainly based on Rellich's lemma and the Stratton--Chu formula for radiating solutions to the Maxwell equations.

math.AP↗

arXiv2019

Multi-objects Generation with Amortized Structural Regularization

Deep generative models (DGMs) have shown promise in image generation. However, most of the existing work learn the model by simply optimizing a divergence between the marginal distributions of the model and the data, and often fail to capture the rich structures and relations in multi-object images. Human knowledge is a critical element to the success of DGMs to infer these structures. In this paper, we propose the amortized structural regularization (ASR) framework, which adopts the posterior regularization (PR) to embed human knowledge into DGMs via a set of structural constraints. We derive a lower bound of the regularized log-likelihood, which can be jointly optimized with respect to the generative model and recognition model efficiently. Empirical results show that ASR significantly outperforms the DGM baselines in terms of inference accuracy and sample quality.

cs.LG↗

arXiv2019

Uniqueness in inverse acoustic and electromagnetic scattering with phaseless near-field data at a fixed frequency

This paper is concerned with uniqueness results in inverse acoustic and electromagnetic scattering problems with phaseless total-field data at a fixed frequency. Motivated by our previous work ({\em SIAM J. Appl. Math. \bf78} (2018), 1737-1753), where uniqueness results were proved for inverse acoustic scattering with phaseless far-field data generated by superpositions of two plane waves as the incident waves at a fixed frequency, in this paper, we use superpositions of two point sources as the incident fields at a fixed frequency and measure the modulus of the acoustic total-field (called phaseless acoustic near-field data) on two spheres enclosing the scatterers generated by such incident fields on the two spheres. Based on this idea, we prove that the impenetrable bounded obstacle or the index of refraction of an inhomogeneous medium can be uniquely determined from the phaseless acoustic near-field data at a fixed frequency. Moreover, the idea is also extended to the electromagnetic case, and it is proved that the impenetrable bounded obstacle or the index of refraction of an inhomogeneous medium can be uniquely determined by the phaseless electric near-field data at a fixed frequency, that is, the modulus of the tangential component with the orientations $\mathbf e_ϕ$ and $\mathbf e_θ$, respectively, of the electric total-field measured on a sphere enclosing the scatters and generated by superpositions of two electric dipoles at a fixed frequency located on the measurement sphere and another bigger sphere with the polarization vectors $\mathbf e_ϕ$ and $\mathbf e_θ$, respectively. As far as we know, this is the first uniqueness result for three-dimensional inverse electromagnetic scattering with phaseless near-field data.

math.AP↗

arXiv2019

Function Space Particle Optimization for Bayesian Neural Networks

While Bayesian neural networks (BNNs) have drawn increasing attention, their posterior inference remains challenging, due to the high-dimensional and over-parameterized nature. To address this issue, several highly flexible and scalable variational inference procedures based on the idea of particle optimization have been proposed. These methods directly optimize a set of particles to approximate the target posterior. However, their application to BNNs often yields sub-optimal performance, as such methods have a particular failure mode on over-parameterized models. In this paper, we propose to solve this issue by performing particle optimization directly in the space of regression functions. We demonstrate through extensive experiments that our method successfully overcomes this issue, and outperforms strong baselines in a variety of tasks including prediction, defense against adversarial examples, and reinforcement learning.

stat.ML↗