arXiv subjects

Bo Xin

Publications and source records attributed to Bo Xin.

18 recordsLinked to original sources

arXiv2021

Mitigating Satellite Trails: a Study of Residual Light after Masking

Using Hyper Suprime Camera data (a precursor of what is to come with Rubin Observatory) we assess trail masking mitigation strategies for satellite contamination. We examine HSC data of the Hubble COSMOS field where satellite trails have been identified by eye. Exercising the current LSST Science Pipelines on this data, we study the efficacy of masking satellite trails which appear in single visit exposures before they are assembled into a coadded frame. We find that the current routines largely mask satellite trails in single visits, but miss the extended low surface brightness features of the satellite trails. For a sufficiently wide mask, these faint features appear at a less significant level in the final coadd, as they are averaged down in a stack of tens of exposures. We study this print-through vs mask width. In this note, we describe some of the challenges we encountered in that effort, prospects for more complete removal of the low surface brightness tails of the masked trails, and possible science impacts.

astro-ph.IM↗

arXiv2020

Improving Person Re-identification with Iterative Impression Aggregation

Our impression about one person often updates after we see more aspects of him/her and this process keeps iterating given more meetings. We formulate such an intuition into the problem of person re-identification (re-ID), where the representation of a query (probe) image is iteratively updated with new information from the candidates in the gallery. Specifically, we propose a simple attentional aggregation formulation to instantiate this idea and showcase that such a pipeline achieves competitive performance on standard benchmarks including CUHK03, Market-1501 and DukeMTMC. Not only does such a simple method improve the performance of the baseline models, it also achieves comparable performance with latest advanced re-ranking methods. Another advantage of this proposal is its flexibility to incorporate different representations and similarity metrics. By utilizing stronger representations and metrics, we further demonstrate state-of-the-art person re-ID performance, which also validates the general applicability of the proposed method.

cs.CV↗

arXiv2020

Mitigation of LEO Satellite Brightness and Trail Effects on the Rubin Observatory LSST

We report studies on the mitigation of optical effects of bright low-Earth-orbit (LEO) satellites on Vera C. Rubin Observatory and its Legacy Survey of Space and Time (LSST). These include options for pointing the telescope to avoid satellites, laboratory investigations of bright trails on the Rubin Observatory LSST camera sensors, algorithms for correcting image artifacts caused by bright trails, experiments on darkening SpaceX Starlink satellites, and ground-based follow-up observations. The original Starlink v0.9 satellites are g ~ 4.5 mag, and the initial experiment "DarkSat" is g ~ 6.1 mag. Future Starlink darkening plans may reach g ~ 7 mag, a brightness level that enables nonlinear image artifact correction to well below background noise. However, the satellite trails will still exist at a signal-to-noise ratio ~ 100, generating systematic errors that may impact data analysis and limit some science. For the Rubin Observatory 8.4-m mirror and a satellite at 550 km, the full width at half maximum of the trail is about 3" as the result of an out-of-focus effect, which helps avoid saturation by decreasing the peak surface brightness of the trail. For 48,000 LEOsats of apparent magnitude 4.5, about 1% of pixels in LSST nautical twilight images would need to be masked.

astro-ph.IM↗

arXiv2020

The LSST DESC Data Challenge 1: Generation and Analysis of Synthetic Images for Next Generation Surveys

Data Challenge 1 (DC1) is the first synthetic dataset produced by the Rubin Observatory Legacy Survey of Space and Time (LSST) Dark Energy Science Collaboration (DESC). DC1 is designed to develop and validate data reduction and analysis and to study the impact of systematic effects that will affect the LSST dataset. DC1 is comprised of $r$-band observations of 40 deg$^{2}$ to 10-year LSST depth. We present each stage of the simulation and analysis process: a) generation, by synthesizing sources from cosmological N-body simulations in individual sensor-visit images with different observing conditions; b) reduction using a development version of the LSST Science Pipelines; and c) matching to the input cosmological catalog for validation and testing. We verify that testable LSST requirements pass within the fidelity of DC1. We establish a selection procedure that produces a sufficiently clean extragalactic sample for clustering analyses and we discuss residual sample contamination, including contributions from inefficiency in star-galaxy separation and imperfect deblending. We compute the galaxy power spectrum on the simulated field and conclude that: i) survey properties have an impact of 50\% of the statistical uncertainty for the scales and models used in DC1 ii) a selection to eliminate artifacts in the catalogs is necessary to avoid biases in the measured clustering; iii) the presence of bright objects has a significant impact (2- to 6-$\sigma$) in the estimated power spectra at small scales ($\ell > 1200$), highlighting the impact of blending in studies at small angular scales in LSST;

astro-ph.IM↗

arXiv2019

Future Uses of the LSST Facility: Input from the LSST Project Science Team

In this white paper, we discuss future uses of the LSST facility after the planned 10-year survey is complete. We expect the LSST survey to profoundly affect the scientific landscape over the next ten years, and it is likely that unexpected discoveries may drive its future scientific program. We discuss various operations and instrument options that could be considered for an extended LSST mission beyond ten years.

astro-ph.IM↗

arXiv2018

A Study of the Point Spread Function in SDSS Images

We use SDSS imaging data in $ugriz$ passbands to study the shape of the point spread function (PSF) profile and the variation of its width with wavelength and time. We find that the PSF profile is well described by theoretical predictions based on von K\'{a}rm\'{a}n's turbulence theory. The observed PSF radial profile can be parametrized by only two parameters, the profile's full width at half maximum (FWHM) and a normalization of the contribution of an empirically determined "instrumental" PSF. The profile shape is very similar to the "double gaussian plus power-law wing" decomposition used by SDSS image processing pipeline, but here it is successfully modeled with two free model parameters, rather than six as in SDSS pipeline. The FWHM variation with wavelength follows the $\lambda^{\alpha}$ power law, where $\alpha \approx-0.3$ and is correlated with the FWHM itself. The observed behavior is much better described by von K\'{a}rm\'{a}n's turbulence theory with the outer scale parameter in the range 5$-$100 m, than by the Kolmogorov's turbulence theory. We also measure the temporal and angular structure functions for FWHM and compare them to simulations and results from literature. The angular structure function saturates at scales beyond 0.5$-$1.0 degree. The power spectrum of the temporal behavior is found to be broadly consistent with a damped random walk model with characteristic timescale in the range $\sim5-30$ minutes, though data show a shallower high-frequency behavior. The latter is well fit by a single power law with index in the range $-1.5$ to $-1.0$. A hybrid model is likely needed to fully capture both the low-frequency and high-frequency behavior of the temporal variations of atmospheric seeing.

astro-ph.IM↗

arXiv2017

Revisiting the Master-Slave Architecture in Multi-Agent Deep Reinforcement Learning

Many tasks in artificial intelligence require the collaboration of multiple agents. We exam deep reinforcement learning for multi-agent domains. Recent research efforts often take the form of two seemingly conflicting perspectives, the decentralized perspective, where each agent is supposed to have its own controller; and the centralized perspective, where one assumes there is a larger model controlling all agents. In this regard, we revisit the idea of the master-slave architecture by incorporating both perspectives within one framework. Such a hierarchical structure naturally leverages advantages from one another. The idea of combining both perspectives is intuitive and can be well motivated from many real world systems, however, out of a variety of possible realizations, we highlights three key ingredients, i.e. composed action representation, learnable communication and independent reasoning. With network designs to facilitate these explicitly, our proposal consistently outperforms latest competing methods both in synthetic experiments and when applied to challenging StarCraft micromanagement tasks.

cs.AI↗

arXiv2017

From Bayesian Sparsity to Gated Recurrent Nets

The iterations of many first-order algorithms, when applied to minimizing common regularized regression functions, often resemble neural network layers with pre-specified weights. This observation has prompted the development of learning-based approaches that purport to replace these iterations with enhanced surrogates forged as DNN models from available training data. For example, important NP-hard sparse estimation problems have recently benefitted from this genre of upgrade, with simple feedforward or recurrent networks ousting proximal gradient-based iterations. Analogously, this paper demonstrates that more powerful Bayesian algorithms for promoting sparsity, which rely on complex multi-loop majorization-minimization techniques, mirror the structure of more sophisticated long short-term memory (LSTM) networks, or alternative gated feedback networks previously designed for sequence prediction. As part of this development, we examine the parallels between latent variable trajectories operating across multiple time-scales during optimization, and the activations within deep network structures designed to adaptively model such characteristic sequences. The resulting insights lead to a novel sparse estimation system that, when granted training data, can estimate optimal solutions efficiently in regimes where other algorithms fail, including practical direction-of-arrival (DOA) and 3D geometry recovery problems. The underlying principles we expose are also suggestive of a learning process for a richer class of multi-loop algorithms in other domains.

cs.LG↗

arXiv2017

Collaborative Deep Reinforcement Learning for Joint Object Search

We examine the problem of joint top-down active search of multiple objects under interaction, e.g., person riding a bicycle, cups held by the table, etc.. Such objects under interaction often can provide contextual cues to each other to facilitate more efficient search. By treating each detector as an agent, we present the first collaborative multi-agent deep reinforcement learning algorithm to learn the optimal policy for joint active object localization, which effectively exploits such beneficial contextual information. We learn inter-agent communication through cross connections with gates between the Q-networks, which is facilitated by a novel multi-agent deep Q-learning algorithm with joint exploitation sampling. We verify our proposed method on multiple object detection benchmarks. Not only does our model help to improve the performance of state-of-the-art active localization models, it also reveals interesting co-detection patterns that are intuitively interpretable.

cs.CV↗

arXiv2016

Maximal Sparsity with Deep Networks?

The iterations of many sparse estimation algorithms are comprised of a fixed linear filter cascaded with a thresholding nonlinearity, which collectively resemble a typical neural network layer. Consequently, a lengthy sequence of algorithm iterations can be viewed as a deep network with shared, hand-crafted layer weights. It is therefore quite natural to examine the degree to which a learned network model might act as a viable surrogate for traditional sparse estimation in domains where ample training data is available. While the possibility of a reduced computational budget is readily apparent when a ceiling is imposed on the number of layers, our work primarily focuses on estimation accuracy. In particular, it is well-known that when a signal dictionary has coherent columns, as quantified by a large RIP constant, then most tractable iterative algorithms are unable to find maximally sparse representations. In contrast, we demonstrate both theoretically and empirically the potential for a trained deep network to recover minimal $\ell_0$-norm representations in regimes where existing methods fail. The resulting system is deployed on a practical photometric stereo estimation problem, where the goal is to remove sparse outliers that can disrupt the estimation of surface normals from a 3D scene.

cs.LG↗

arXiv2015

Curvature Wavefront Sensing for the Large Synoptic Survey Telescope

The Large Synoptic Survey Telescope (LSST) will use an active optics system (AOS) to maintain alignment and surface figure on its three large mirrors. Corrective actions fed to the LSST AOS are determined from information derived from 4 curvature wavefront sensors located at the corners of the focal plane. Each wavefront sensor is a split detector such that the halves are 1mm on either side of focus. In this paper we describe the extensions to published curvature wavefront sensing algorithms needed to address challenges presented by the LSST, namely the large central obscuration, the fast f/1.23 beam, off-axis pupil distortions, and vignetting at the sensor locations. We also describe corrections needed for the split sensors and the effects from the angular separation of different stars providing the intra- and extra-focal images. Lastly, we present simulations that demonstrate convergence, linearity, and negligible noise when compared to atmospheric effects when the algorithm extensions are applied to the LSST optical system. The algorithm extensions reported here are generic and can easily be adapted to other wide-field optical systems including similar telescopes with large central obscuration and off-axis curvature sensing.

astro-ph.IM↗

arXiv2015

Background Subtraction via Generalized Fused Lasso Foreground Modeling

Background Subtraction (BS) is one of the key steps in video analysis. Many background models have been proposed and achieved promising performance on public data sets. However, due to challenges such as illumination change, dynamic background etc. the resulted foreground segmentation often consists of holes as well as background noise. In this regard, we consider generalized fused lasso regularization to quest for intact structured foregrounds. Together with certain assumptions about the background, such as the low-rank assumption or the sparse-composition assumption (depending on whether pure background frames are provided), we formulate BS as a matrix decomposition problem using regularization terms for both the foreground and background matrices. Moreover, under the proposed formulation, the two generally distinctive background assumptions can be solved in a unified manner. The optimization was carried out via applying the augmented Lagrange multiplier (ALM) method in such a way that a fast parametric-flow algorithm is used for updating the foreground matrix. Experimental results on several popular BS data sets demonstrate the advantage of the proposed model compared to state-of-the-arts.

cs.CV↗

arXiv2015

Stable Feature Selection from Brain sMRI

Neuroimage analysis usually involves learning thousands or even millions of variables using only a limited number of samples. In this regard, sparse models, e.g. the lasso, are applied to select the optimal features and achieve high diagnosis accuracy. The lasso, however, usually results in independent unstable features. Stability, a manifest of reproducibility of statistical results subject to reasonable perturbations to data and the model, is an important focus in statistics, especially in the analysis of high dimensional data. In this paper, we explore a nonnegative generalized fused lasso model for stable feature selection in the diagnosis of Alzheimer's disease. In addition to sparsity, our model incorporates two important pathological priors: the spatial cohesion of lesion voxels and the positive correlation between the features and the disease labels. To optimize the model, we propose an efficient algorithm by proving a novel link between total variation and fast network flow algorithms via conic duality. Experiments show that the proposed nonnegative model performs much better in exploring the intrinsic structure of data via selecting stable features compared with other state-of-the-arts.

cs.LG↗

arXiv2014

Exploring Algorithmic Limits of Matrix Rank Minimization under Affine Constraints

Many applications require recovering a matrix of minimal rank within an affine constraint set, with matrix completion a notable special case. Because the problem is NP-hard in general, it is common to replace the matrix rank with the nuclear norm, which acts as a convenient convex surrogate. While elegant theoretical conditions elucidate when this replacement is likely to be successful, they are highly restrictive and convex algorithms fail when the ambient rank is too high or when the constraint set is poorly structured. Non-convex alternatives fare somewhat better when carefully tuned; however, convergence to locally optimal solutions remains a continuing source of failure. Against this backdrop we derive a deceptively simple and parameter-free probabilistic PCA-like algorithm that is capable, over a wide battery of empirical tests, of successful recovery even at the theoretical limit where the number of measurements equal the degrees of freedom in the unknown low-rank matrix. Somewhat surprisingly, this is possible even when the affine constraint set is highly ill-conditioned. While proving general recovery guarantees remains evasive for non-convex algorithms, Bayesian-inspired or otherwise, we nonetheless show conditions whereby the underlying cost function has a unique stationary point located at the global optimum; no existing cost function we are aware of satisfies this same property. We conclude with a simple computer vision application involving image rectification and a standard collaborative filtering benchmark.

cs.LG↗

arXiv2012

Fast Shape Estimation for Galaxies and Stars

Model fitting is frequently used to determine the shape of galaxies and the point spread function, for examples, in weak lensing analyses or morphology studies aiming at probing the evolution of galaxies. However, the number of parameters in the model, as well as the number of objects, are often so large as to limit the use of model fitting for future large surveys. In this article, we propose a set of algorithms to speed up the fitting process. Our approach is divided into three distinctive steps: centroiding, ellipticity measurement, and profile fitting. We demonstrate that we can derive the position and ellipticity of an object analytically in the first two steps and thus leave only a small number of parameters to be derived through model fitting. The position, ellipticity, and shape parameters can then used in constructing orthonomal basis functions such as sérsiclets for better galaxy image reconstruction. We assess the efficiency and accuracy of the algorithms with simulated images. We have not taken into account the deconvolution of the point spread function, which most weak lensing analyses do.

astro-ph.CO↗

arXiv2011

Charm Semileptonic Decays at CLEO-c

We review the recent results on D and Ds meson semileptonic decays from CLEO-c. Comparisons with lattice quantum chromodynamics (LQCD) calculations and implications for B physics are also discussed.

hep-ex↗

arXiv2011

Leptonic Charm Decays

We review the recent experimental results on D and Ds meson leptonic decays from CLEO-c, Belle, and BABAR, which results in the decay constants f_D= (206.7 +- 8.9) MeV and f_Ds= (257.3 +- 5.3) MeV. The latter is an average obtained by the Heavy Flavor Averaging Group (HFAG). Comparisons with Lattice QCD (LQCD) calculations are discussed.

hep-ex↗

arXiv2009

Recent Results on Charm Semileptonic Decays

We review the recent results on D and D_s meson semileptonic decays from CLEO-c, BABAR, and Belle. Comparisons with lattice quantum chromodynamics (LQCD) calculations and implications for B physics are also discussed.

hep-ex↗