arXiv subjects

Bin Liu

Publications and source records attributed to Bin Liu.

At least 325 records · Page 18Linked to original sources

arXiv2022

Walle: An End-to-End, General-Purpose, and Large-Scale Production System for Device-Cloud Collaborative Machine Learning

To break the bottlenecks of mainstream cloud-based machine learning (ML) paradigm, we adopt device-cloud collaborative ML and build the first end-to-end and general-purpose system, called Walle, as the foundation. Walle consists of a deployment platform, distributing ML tasks to billion-scale devices in time; a data pipeline, efficiently preparing task input; and a compute container, providing a cross-platform and high-performance execution environment, while facilitating daily task iteration. Specifically, the compute container is based on Mobile Neural Network (MNN), a tensor compute engine along with the data processing and model execution libraries, which are exposed through a refined Python thread-level virtual machine (VM) to support diverse ML tasks and concurrent task execution. The core of MNN is the novel mechanisms of operator decomposition and semi-auto search, sharply reducing the workload in manually optimizing hundreds of operators for tens of hardware backends and further quickly identifying the best backend with runtime optimization for a computation graph. The data pipeline introduces an on-device stream processing framework to enable processing user behavior data at source. The deployment platform releases ML tasks with an efficient push-then-pull method and supports multi-granularity deployment policies. We evaluate Walle in practical e-commerce application scenarios to demonstrate its effectiveness, efficiency, and scalability. Extensive micro-benchmarks also highlight the superior performance of MNN and the Python thread-level VM. Walle has been in large-scale production use in Alibaba, while MNN has been open source with a broad impact in the community.

cs.LG↗

arXiv2022

Emergence of Double-slit Interference by Representing Visual Space in Artificial Neural Networks

Artificial neural networks have realized incredible successes at image recognition, but the underlying mechanism of visual space representation remains a huge mystery. Grid cells (2014 Nobel Prize) in the entorhinal cortex support a periodic representation as a metric for coding space. Here, we develop a self-supervised convolutional neural network to perform visual space location, leading to the emergence of single-slit diffraction and double-slit interference patterns of waves. Our discoveries reveal the nature of CNN encoding visual space to a certain extent. CNN is no longer a black box in terms of visual spatial encoding, it is interpretable. Our findings indicate that the periodicity property of waves provides a space metric, suggesting a general role of spatial coordinate frame in artificial neural networks.

cs.CV↗

arXiv2022

Massive molecular gas reservoir in a luminous sub-millimeter galaxy during cosmic noon

We present multi-band observations of an extremely dusty star-forming lensed galaxy (HERS1) at $z=2.553$. High-resolution maps of \textit{HST}/WFC3, SMA, and ALMA show a partial Einstein-ring with a radius of $\sim$3$^{\prime\prime}$. The deeper HST observations also show the presence of a lensing arc feature associated with a second lens source, identified to be at the same redshift as the bright arc based on a detection of the [NII] 205$μ$m emission line with ALMA. A detailed model of the lensing system is constructed using the high-resolution HST/WFC3 image, which allows us to study the source plane properties and connect rest-frame optical emission with properties of the galaxy as seen in sub-millimeter and millimeter wavelengths. Corrected for lensing magnification, the spectral energy distribution fitting results yield an intrinsic star formation rate of about $1000\pm260$ ${\rm M_{\odot}}$yr$^{-1}$, a stellar mass ${\rm M_*}=4.3^{+2.2}_{-1.0}\times10^{11} {\rm M_{\odot}}$, and a dust temperature ${\rm T}_{\rm d}=35^{+2}_{-1}$ K. The intrinsic CO emission line ($J_{\rm up}=3,4,5,6,7,9$) flux densities and CO spectral line energy distribution are derived based on the velocity-dependent magnification factors. We apply a radiative transfer model using the large velocity gradient method with two excitation components to study the gas properties. The low-excitation component has a gas density $n_{\rm H_2}=10^{3.1\pm0.6}$ cm$^{-3}$ and kinetic temperature ${\rm T}_{\rm k}=19^{+7}_{-5}$ K and a high-excitation component has $n_{\rm H_2}=10^{2.8\pm0.3}$ cm$^{-3}$ and ${\rm T}_{\rm k}=550^{+260}_{-220}$ K. Additionally, HERS1 has a gas fraction of about $0.4\pm0.2$ and is expected to last 250 Myr. These properties offer a detailed view of a typical sub-millimeter galaxy during the peak epoch of star-formation activity.

astro-ph.GA↗

arXiv2022

Evolution of Stellar Orbits Around Merging Massive Black-Hole Binary

We study the long-term orbital evolution of stars around a merging massive or supermassive black-hole (BH) binary, taking into account the general relativistic effect induced by the BH spin. When the BH spin is significant compared to and misaligned with the binary orbital angular momentum, the orbital axis ($\hat{\mathbf{l}}$) of the circumbinary star can undergo significant evolution during the binary orbital decay driven by gravitational radiation. Including the spin effect of the primary (more massive) BH, we find that starting from nearly coplanar orbital orientations, the orbital axes $\hat{\mathbf{l}}$ of circumbinary stars preferentially evolve towards the spin direction after the merger of the BH binary, regardless of the initial BH spin orientation. Such alignment phenomenon, i.e., small final misalignment angle between $\hat{\mathbf{l}}$ and the spin axis of the remanent BH $\hat{\mathbf{S}}$, can be understood analytically using the principle of adiabatic invariance. For the BH binaries with extremely mass ratio ($m_2/m_1\lesssim0.01$), $\hat{\mathbf{l}}$ may experience more complicated evolution as adiabatic invariance breaks down, but the trend of alignment still works reasonably well when the initial binary spin-orbit angle is relatively small. Our result suggests that the correlation between the orientations of stellar orbits and the spin axis of the central BH could provide a potential signature of the merger history of the massive BH.

astro-ph.HE↗

arXiv2022

The FAST Ultra-Deep Survey (FUDS): observational strategy, calibration and data reduction

The FAST Ultra-Deep Survey (FUDS) is a blind survey that aims for the direct detection of HI in galaxies at redshifts $z<0.42$. The survey uses the multibeam receiver on the Five Hundred Meter Aperture Spherical Telescope (FAST) to map six regions, each of size 0.72 deg$^2$ at high sensitivity ($\sim 50 μ$Jy) and high frequency resolution (23 kHz). The survey will enable studies of the evolution of galaxies and their HI content with an eventual sample size of $\sim 1000$. We present the science goals, observing strategy, the effects of radio frequency interference (RFI) at the FAST site, our mitigation strategies and the methods for calibration, data reduction and imaging as applied to initial data. The observations and reductions for the first field, FUDS0, are completed, with around 128 HI galaxies detected in a preliminary analysis. Example spectra are given in this paper, including a comparison with data from the overlapping GAL2577 field of Arecibo Ultra-Deep Survey (AUDS).

astro-ph.GA↗

arXiv2022

Real-time Online Multi-Object Tracking in Compressed Domain

Recent online Multi-Object Tracking (MOT) methods have achieved desirable tracking performance. However, the tracking speed of most existing methods is rather slow. Inspired from the fact that the adjacent frames are highly relevant and redundant, we divide the frames into key and non-key frames respectively and track objects in the compressed domain. For the key frames, the RGB images are restored for detection and data association. To make data association more reliable, an appearance Convolutional Neural Network (CNN) which can be jointly trained with the detector is proposed. For the non-key frames, the objects are directly propagated by a tracking CNN based on the motion information provided in the compressed domain. Compared with the state-of-the-art online MOT methods,our tracker is about 6x faster while maintaining a comparable tracking performance.

cs.CV↗

arXiv2022

Experimental Demonstration of High-Rate Discrete-Modulated Continuous-Variable Quantum Key Distribution System

A high-rate continuous-variable quantum key distribution (CV-QKD) system based on high-order discrete modulation is experimentally investigated. With the help of the novel system scheme, effective digital signal processing algorithms and advanced analytical security proof method, the transmission results of 5 km, 10 km, 25 km, and 50 km are achieved for the 1 GBaud optimized quantum signals. Correspondingly, the asymptotic secret key rate (SKR) is 288.421 Mbps, 159.395 Mbps, 50.004 Mbps and 7.579 Mbps for discrete Gaussian (DG) 64QAM, and 326.708 Mbps, 179.348 Mbps, 50.623 Mbps and 9.212 Mbps for DG 256QAM. Under the same parameters, the achieved SKRs of DG 256QAM is almost same to ideal Gaussian modulation. In this case, the demonstrated high-rate discrete modulated CV-QKD system has the application potential for high speed security communication under tens of kilometers.

quant-ph↗

arXiv2022

Notes on BIM and BFM Optimal Power Flow With Parallel Lines and Total Current Limits

The second-order cone relaxation of the branch flow model (BFM) and bus injection model (BIM) variants of optimal power flow are well-known to be equivalent for radial networks. In this work we show that in meshed networks with parallel lines, BIM dominates BFM, and propose novel constraints to make them equivalent in general. Furthermore, we develop an improvement to the second-order cone relaxations of optimal power flow, adding novel and valid linear constraints on the lifted current expressions. We develop two simple test cases to highlight the advantages of the proposed constraints. These novel constraints tighten the second-order cone relaxation gap on test cases in the `PG Lib' optimal power flow benchmark library, albeit generally in limited fashion.

math.OC↗

arXiv2022

Approximate Nearest Neighbor Search under Neural Similarity Metric for Large-Scale Recommendation

Model-based methods for recommender systems have been studied extensively for years. Modern recommender systems usually resort to 1) representation learning models which define user-item preference as the distance between their embedding representations, and 2) embedding-based Approximate Nearest Neighbor (ANN) search to tackle the efficiency problem introduced by large-scale corpus. While providing efficient retrieval, the embedding-based retrieval pattern also limits the model capacity since the form of user-item preference measure is restricted to the distance between their embedding representations. However, for other more precise user-item preference measures, e.g., preference scores directly derived from a deep neural network, they are computationally intractable because of the lack of an efficient retrieval method, and an exhaustive search for all user-item pairs is impractical. In this paper, we propose a novel method to extend ANN search to arbitrary matching functions, e.g., a deep neural network. Our main idea is to perform a greedy walk with a matching function in a similarity graph constructed from all items. To solve the problem that the similarity measures of graph construction and user-item matching function are heterogeneous, we propose a pluggable adversarial training task to ensure the graph search with arbitrary matching function can achieve fairly high precision. Experimental results in both open source and industry datasets demonstrate the effectiveness of our method. The proposed method has been fully deployed in the Taobao display advertising platform and brings a considerable advertising revenue increase. We also summarize our detailed experiences in deployment in this paper.

cs.IR↗

arXiv2022

WSLRec: Weakly Supervised Learning for Neural Sequential Recommendation Models

Learning the user-item relevance hidden in implicit feedback data plays an important role in modern recommender systems. Neural sequential recommendation models, which formulates learning the user-item relevance as a sequential classification problem to distinguish items in future behaviors from others based on the user's historical behaviors, have attracted a lot of interest in both industry and academic due to their substantial practical value. Though achieving many practical successes, we argue that the intrinsic {\bf incompleteness} and {\bf inaccuracy} of user behaviors in implicit feedback data is ignored and conduct preliminary experiments for supporting our claims. Motivated by the observation that model-free methods like behavioral retargeting (BR) and item-based collaborative filtering (ItemCF) hit different parts of the user-item relevance compared to neural sequential recommendation models, we propose a novel model-agnostic training approach called WSLRec, which adopts a three-stage framework: pre-training, top-$k$ mining, and fine-tuning. WSLRec resolves the incompleteness problem by pre-training models on extra weak supervisions from model-free methods like BR and ItemCF, while resolves the inaccuracy problem by leveraging the top-$k$ mining to screen out reliable user-item relevance from weak supervisions for fine-tuning. Experiments on two benchmark datasets and online A/B tests verify the rationality of our claims and demonstrate the effectiveness of WSLRec.

cs.IR↗

arXiv2022

Concise and Effective Network for 3D Human Modeling from Orthogonal Silhouettes

In this paper, we revisit the problem of 3D human modeling from two orthogonal silhouettes of individuals (i.e., front and side views). Different from our prior work, a supervised learning approach based on convolutional neural network (CNN) is investigated to solve the problem by establishing a mapping function that can effectively extract features from two silhouettes and fuse them into coefficients in the shape space of human bodies. A new CNN structure is proposed in our work to exact not only the discriminative features of front and side views and also their mixed features for the mapping function. 3D human models with high accuracy are synthesized from coefficients generated by the mapping function. Existing CNN approaches for 3D human modeling usually learn a large number of parameters (from 8.5M to 355.4M) from two binary images. Differently, we investigate a new network architecture and conduct the samples on silhouettes as input. As a consequence, more accurate models can be generated by our network with only 2.4M coefficients. The training of our network is conducted on samples obtained by augmenting a publicly accessible dataset. Learning transfer by using datasets with a smaller number of scanned models is applied to our network to enable the function of generating results with gender-oriented (or geographical) patterns.

cs.CV↗

arXiv2022

Load Balancing in Low-Voltage Distribution Networks via Optimizing Residential Phase Connections

Unbalance issues in low-voltage distribution networks (LVDN) can be worsened by increasing penetration of residential PV generation if unevenly distributed among three phases. To address this issue, the phase-switching device (PSD) provides a viable and efficient method by dynamically switching customers to other phases. This paper further investigates how to optimize residential phase connections by controlling PSDs efficiently. The optimization problem is formulated as a mixed-integer non-convex programming (MINCP) problem considering relevant operational requirements of an LVDN based on the exact formulation of unbalanced three-phase optimal power flow (UTOPF). Unlike most heuristic algorithms and the linearization techniques in our previous work, this paper proposes to solve the MINCP problem via an iteration-based algorithm after exact reformulations and reasonable approximations of some constraints. The proposed method is tested in a real LVDN and compared with the approach of Zhao et al. based on the well-known linear UTOPF formulation. Case studies based on the European low-voltage test feeder demonstrate the proposed method's efficiency in mitigating the network unbalance while ensuring network security and flexibility to deal with more controllable resources.

math.OC↗

arXiv2022

Thermalization of fluorescent protein exciton-polaritons at room temperature

Fluorescent proteins (FPs) have recently emerged as a serious contender for realizing ultralow threshold room temperature exciton-polariton condensation and lasing. Our contribution investigates the thermalization of FP microcavity exciton-polaritons upon optical pumping under ambient conditions. We realize polariton cooling using a new FP molecule, called mScarlet, coupled strongly to the optical modes in a Fabry Perot cavity. Interestingly, at the threshold excitation energy (fluence) of ~ 9 nJ/pulse (15.6 mJ/cm2), we observe an effective temperature, Teff ~ 350 +/- 35 K close to the lattice temperature indicative of strongly thermalized exciton-polaritons at equilibrium. This efficient thermalization results from the interplay of radiative pumping facilitated by the energetics of the lower polariton branch and the cavity Q factor. Direct evidence for dramatic switching from an equilibrium state into a metastable state is observed for the organic cavity polariton device at room temperature via deviation from the Maxwell-Boltzmann statistics at k = 0 above the threshold. Thermalized polariton gases in organic systems at equilibrium hold substantial promise for designing room temperature polaritonic circuits, switches, and lattices for analog simulation.

physics.optics↗

arXiv2022

Probing the Spins of Supermassive Black Holes with Gravitational Waves from Surrounding Compact Binaries

Merging compact black-hole (BH) binaries are likely to exist in the nuclear star clusters around supermassive BHs (SMBHs), such as Sgr A$^\ast$. They may also form in the accretion disks of active galactic nuclei. Such compact binaries can emit gravitational waves (GWs) in the low-frequency band (0.001-1 Hz) that are detectable by several planned space-borne GW observatories. We show that the orbital axis of the compact binary may experience significant variation due to the frame-dragging effect associated with the spin of the SMBH. The dynamical behavior of the orbital axis can be understood analytically as a resonance phenomenon. We show that rate of change of the binary orbital axis encodes the information on the spin of the SMBH. Therefore detecting GWs from compact binaries around SMBHs, particularly the modulation of the waveform associated with the variation of the binary orbital axis, can provide a new probe on the spins of SMBHs.

astro-ph.HE↗

arXiv2022

Optimizing Area Under the Curve Measures via Matrix Factorization for Predicting Drug-Target Interaction with Multiple Similarities

In drug discovery, identifying drug-target interactions (DTIs) via experimental approaches is a tedious and expensive procedure. Computational methods efficiently predict DTIs and recommend a small part of potential interacting pairs for further experimental confirmation, accelerating the drug discovery process. Although it has been shown that fusing heterogeneous drug and target similarities can improve the prediction ability, the existing similarity combination methods ignore the interaction consistency for neighbour entities which is more crucial for the DTI prediction model. Furthermore, area under the precision-recall curve (AUPR) that emphasizes the accuracy of top-ranked pairs and area under the receiver operating characteristic curve (AUC) that heavily punishes the existence of low ranked interacting pairs are two widely used evaluation metrics in DTI prediction. However, the two metrics are seldom considered as losses within existing DTI prediction methods. This paper first proposes two matrix factorization (MF) methods that optimize AUPR and AUC using convex surrogate losses respectively, and then develops an ensemble MF approach takes advantage of the two area under the curve metrics by combining the two single metric based MF models. Both three proposed approaches incorporate a novel local interaction consistency aware similarity interaction method to generate fused drug and target similarities that preserve vital information from the more reliable view. Experimental results over five datasets under different prediction settings show that the proposed methods outperform various competitors in terms of the metric(s) they optimize. In addition, the validation on the top ranked novel predictions confirms the ability of our methods to discover potential new DTIs.

q-bio.QM↗

arXiv2022

Numerical study of three-dimensional single-mode Rayleigh-Taylor instability in turbulent mixing stage

Rayleigh-Taylor instability (RTI) as a multi-scale, strongly nonlinear physical phenomenon which plays an important role in the engineering applications and scientific research. In this paper, the mesoscopic lattice Boltzmann method is used to numerically study the late-time evolutional mechanism of three-dimensional (3D) single-mode RTI and the influences of extensive dimensionless Reynolds number and Atwood number on phase interfacial dynamics, spike and bubble growth are investigated in details. For a high Reynolds number, it is reported that the development of 3D single-mode RTI would undergo four different stages: linear growth stage, saturated velocity growth stage, reacceleration stage and turbulent mixing stage. A series of complex interfacial structures with large topological changes can be observed at the turbulent mixing stage, which always preserve the symmetries with respect to the middle axis at a low Atwood number, and the lines of symmetry within spike and bubble are broken as the Atwood number is increased. Five statistical methods for computing the spike and bubble growth rates are then analyzed to reveal the growth law of 3D single-mode RTI in turbulent mixing stage. It is found that the spike late-time growth rate shows an overall increase with the Atwood number, while the bubble growth rate seems to be independence of the Atwood number, approaching a constant of around 0.1. When the Reynolds number decreases, the later stages cannot be reached gradually and the evolution of phase interface presents a laminar flow state.

physics.flu-dyn↗

arXiv2022

Revealing sign-reversal $s^{+-}$-wave pairing by quasiparticle interference in the heavy-fermion superconductor CeCu$_2$Si$_2$

Recent observations of two nodeless gaps in superconducting CeCu$_2$Si$_2$ have raised intensive debates as to its exact gap structure of either sign-reversal ($s^{+-}$) or sign-preserving ($s^{++}$) pairing. Here we investigate the quasiparticle interference (QPI) using realistic Fermi surface topology for both weak and strong interband impurity scatterings. Our calculations of the QPI and integrated antisymmetrized local density of states reveal qualitative distinctions between $s^{+-}$ and $s^{++}$ pairing states, which include the intragap impurity resonance and a significant energy-dependence difference between two gap energies. Our predictions provide a guide for phase-sensitive QPI measurements to uncover decisively the true pairing symmetry in the heavy-fermion superconductor CeCu$_2$Si$_2$.

cond-mat.supr-con↗

arXiv2022

Online Multi-Object Tracking with Unsupervised Re-Identification Learning and Occlusion Estimation

Occlusion between different objects is a typical challenge in Multi-Object Tracking (MOT), which often leads to inferior tracking results due to the missing detected objects. The common practice in multi-object tracking is re-identifying the missed objects after their reappearance. Though tracking performance can be boosted by the re-identification, the annotation of identity is required to train the model. In addition, such practice of re-identification still can not track those highly occluded objects when they are missed by the detector. In this paper, we focus on online multi-object tracking and design two novel modules, the unsupervised re-identification learning module and the occlusion estimation module, to handle these problems. Specifically, the proposed unsupervised re-identification learning module does not require any (pseudo) identity information nor suffer from the scalability issue. The proposed occlusion estimation module tries to predict the locations where occlusions happen, which are used to estimate the positions of missed objects by the detector. Our study shows that, when applied to state-of-the-art MOT methods, the proposed unsupervised re-identification learning is comparable to supervised re-identification learning, and the tracking performance is further improved by the proposed occlusion estimation module.

cs.CV↗