arXiv subjects

Lu Zheng

Publications and source records attributed to Lu Zheng.

25 records · Page 2Linked to original sources

arXiv2018

Collaborative Multi-modal deep learning for the personalized product retrieval in Facebook Marketplace

Facebook Marketplace is quickly gaining momentum among consumers as a favored customer-to-customer (C2C) product trading platform. The recommendation system behind it helps to significantly improve the user experience. Building the recommendation system for Facebook Marketplace is challenging for two reasons: 1) Scalability: the number of products in Facebook Marketplace is huge. Tens of thousands of products need to be scored and recommended within a couple hundred milliseconds for millions of users every day; 2) Cold start: the life span of the C2C products is very short and the user activities on the products are sparse. Thus it is difficult to accumulate enough product level signals for recommendation and we are facing a significant cold start issue. In this paper, we propose to address both the scalability and the cold-start issue by building a collaborative multi-modal deep learning based retrieval system where the compact embeddings for the users and the products are trained with the multi-modal content information. This system shows significant improvement over the benchmark in online and off-line experiments: In the online experiment, it increases the number of messages initiated by the buyer to the seller by +26.95%; in the off-line experiment, it improves the prediction accuracy by +9.58%.

cs.IR↗

arXiv2018

Visualization of Surface-acoustic-wave Potential by Transmission-mode Microwave Impedance Microscopy

Elastic waves propagating in piezoelectric materials are accompanied by a time-varying electric potential, which is of critical importance for acousto-electronic applications. The spatial mapping of such a potential at microwave frequencies is challenging since the characteristic length scale is determined by the acoustic wavelength of several micrometers. In this work, we report the visualization of surface acoustic waves (SAWs) on ferroelectric samples by transmission-mode microwave impedance microscopy (T-MIM). The SAW potential launched by the interdigital transducer is detected by the tip and demodulated by the microwave electronics as time-independent spatial patterns. Wave phenomena such as interference and diffraction are imaged and the results are in excellent agreement with the theoretical analysis. Our work opens up a new avenue to study various electromechanical systems in a spatially resolved manner.

cond-mat.mes-hall↗

arXiv2018

Interferometric Imaging of Nonlocal Electromechanical Power Transduction in Ferroelectric Domains

The electrical generation and detection of elastic waves are the foundation for acousto-electronic and acousto-optic systems. For surface-acoustic-wave devices, micro-/nano-electromechanical systems, and phononic crystals, tailoring the spatial variation of material properties such as piezoelectric and elastic tensors may bring significant improvements to the system performance. Due to the much smaller speed of sound than speed of light in solids, it is desirable to study various electroacoustic behaviors at the mesoscopic length scale. In this work, we demonstrate the interferometric imaging of electromechanical power transduction in ferroelectric lithium niobate domain structures by microwave impedance microscopy. In sharp contrast to the traditional standing-wave patterns caused by the superposition of counter-propagating waves, the constructive and destructive fringes in microwave dissipation images exhibit an intriguing one-wavelength periodicity. We show that such unusual interference patterns, which are fundamentally different from the acoustic displacement fields, stem from the nonlocal interaction between electric fields and elastic waves. The results are corroborated by numerical simulations taking into account the sign reversal of piezoelectric tensor in oppositely polarized domains. Our work paves new ways to probe nanoscale electroacoustic phenomena in complex structures by near-field electromagnetic imaging.

cond-mat.mes-hall↗

arXiv2018

Quantitative Measurements of Nanoscale Permittivity and Conductivity Using Tuning-fork-based Microwave Impedance Microscopy

We report quantitative measurements of nanoscale permittivity and conductivity using tuning-fork (TF) based microwave impedance microscopy (MIM). The system is operated under the driving amplitude modulation mode, which ensures satisfactory feedback stability on samples with rough surfaces. The demodulated MIM signals on a series of bulk dielectrics are in good agreement with results simulated by finite-element analysis. Using the TF-MIM, we have visualized the evolution of nanoscale conductance on back-gated $MoS_2$ field effect transistors and the results are consistent with the transport data. Our work suggests that quantitative analysis of mesoscopic electrical properties can be achieved by near-field microwave imaging with small distance modulation.

physics.app-ph↗

arXiv2017

Low-energy Structural Dynamics of Ferroelectric Domain Walls in Hexagonal Rare-earth Manganites

Domain walls (DWs) in ferroic materials, across which the order parameter abruptly changes its orientation, can host emergent properties that are absent in the bulk domains. Using a broadband ($10^6-10^{10}$ Hz) scanning impedance microscope, we show that the electrical response of the interlocked antiphase boundaries and ferroelectric domain walls in hexagonal rare-earth manganites ($h-RMnO_3$) is dominated by the bound-charge oscillation rather than free-carrier conduction at the DWs. As a measure of the rate of energy dissipation, the effective conductivity of DWs on the (001) surfaces of $h-RMnO_3$ at GHz frequencies is drastically higher than that at dc, while the effect is absent on surfaces with in-plane polarized domains. First-principles and model calculations indicate that the frequency range and selection rules are consistent with the periodic sliding of the DW around its equilibrium position. This acoustic-wave-like mode, which is associated with the synchronized oscillation of local polarization and apical oxygen atoms, is localized perpendicular to the DW but free to propagate along the DW plane. Our results break the ground to understand structural DW dynamics and exploit new interfacial phenomena for novel devices.

cond-mat.mes-hall↗

arXiv2012

Belief Propagation by Message Passing in Junction Trees: Computing Each Message Faster Using GPU Parallelization

Compiling Bayesian networks (BNs) to junction trees and performing belief propagation over them is among the most prominent approaches to computing posteriors in BNs. However, belief propagation over junction tree is known to be computationally intensive in the general case. Its complexity may increase dramatically with the connectivity and state space cardinality of Bayesian network nodes. In this paper, we address this computational challenge using GPU parallelization. We develop data structures and algorithms that extend existing junction tree techniques, and specifically develop a novel approach to computing each belief propagation message in parallel. We implement our approach on an NVIDIA GPU and test it using BNs from several applications. Experimentally, we study how junction tree parameters affect parallelization opportunities and hence the performance of our algorithm. We achieve speedups ranging from 0.68 to 9.18 for the BNs studied.

cs.AI↗

arXiv2008

Multi-center clinical trials: Randomization and ancillary statistics

The purpose of this paper is to investigate and develop methods for analysis of multi-center randomized clinical trials which only rely on the randomization process as a basis of inference. Our motivation is prompted by the fact that most current statistical procedures used in the analysis of randomized multi-center studies are model based. The randomization feature of the trials is usually ignored. An important characteristic of model based analysis is that it is straightforward to model covariates. Nevertheless, in nearly all model based analyses, the effects due to different centers and, in general, the design of the clinical trials are ignored. An alternative to a model based analysis is to have analyses guided by the design of the trial. Our development of design based methods allows the incorporation of centers as well as other features of the trial design. The methods make use of conditioning on the ancillary statistics in the sample space generated by the randomization process. We have investigated the power of the methods and have found that, in the presence of center variation, there is a significant increase in power. The methods have been extended to group sequential trials with similar increases in power.

stat.AP↗