arXiv subjects

Yu Lei

Publications and source records attributed to Yu Lei.

94 records · Page 6Linked to original sources

arXiv2017

Semi-supervised Multimodal Hashing

Retrieving nearest neighbors across correlated data in multiple modalities, such as image-text pairs on Facebook and video-tag pairs on YouTube, has become a challenging task due to the huge amount of data. Multimodal hashing methods that embed data into binary codes can boost the retrieving speed and reduce storage requirement. As unsupervised multimodal hashing methods are usually inferior to supervised ones, while the supervised ones requires too much manually labeled data, the proposed method in this paper utilizes a part of labels to design a semi-supervised multimodal hashing method. It first computes the transformation matrices for data matrices and label matrix. Then, with these transformation matrices, fuzzy logic is introduced to estimate a label matrix for unlabeled data. Finally, it uses the estimated label matrix to learn hashing functions for data in each modality to generate a unified binary code matrix. Experiments show that the proposed semi-supervised method with 50% labels can get a medium performance among the compared supervised ones and achieve an approximate performance to the best supervised method with 90% labels. With only 10% labels, the proposed method can still compete with the worst compared supervised one.

cs.IR↗

arXiv2017

Low temperature synthesis of heterostructures of transition metal dichalcogenide alloys (WxMo1-xS2) and graphene with superior catalytic performance for hydrogen evolution

Large-area ($\sim$cm$^2$) films of vertical heterostructures formed by alternating graphene and transition-metal dichalcogenide(TMD) alloys are obtained by wet chemical routes followed by a thermal treatment at low temperature (300 $^\circ$C). In particular, we synthesized stacked graphene and W$_x$Mo$_{1-x}$S$_2$ alloy phases that were used as hydrogen evolution catalysts. We observed a Tafel slope of 38.7 mV dec$^{-1}$ and 96 mV onset potential (at current density of 10 mA cm$^{-2}$) when the heterostructure alloy is annealed at 300 $^o$C. These results indicate that heterostructure formed by graphene and W$_{0.4}$Mo$_{0.6}$S$_2$ alloys are far more efficient than WS$_2$ and MoS$_2$ by at least a factor of two, and it is superior than any other reported TMD system. This strategy offers a cheap and low temperature synthesis alternative able to replace Pt in the hydrogen evolution reaction (HER). Furthermore, the catalytic activity of the alloy is stable over time, i.e. the catalytic activity does not experience a significant change even after 1000 cycles. Using density functional theory calculations, we found that this enhanced hydrogen evolution in the W$_x$Mo$_{1-x}$S$_2$ alloys is mainly due to the lower energy barrier created by a favorable overlap of the d-orbitals from the transition metals and the s-orbitals of H$_2$, with the lowest energy barrier occurring for W$_{0.4}$Mo$_{0.6}$S$_2$ alloy. Thus, it is now possible to further improve the performance of the "inert" TMD basal plane via metal alloying, in addition to the previously reported strategies of creation of point defects, vacancies and edges. The synthesis of graphene/W$_{0.4}$Mo$_{0.6}$S$_2$ produced at relatively low temperatures is scalable and could be used as an effective low cost Pt-free catalyst.

cond-mat.mtrl-sci↗

arXiv2017

Programmable Molecular Composites of Tandem Proteins with Graphene-Oxide for Efficient Bimoprh Actuators

The rapid expansion in the spectrum of two-dimensional (2D) materials has driven the efforts of research on the fabrication of 2D composites and heterostructures. Highly ordered structure of 2D materials provides an excellent platform for controlling the ultimate structure and properties of the composite material with precision. However, limited control over the structure of the adherent material and its interactions with highly ordered 2D materials results in defective composites with inferior performance. Here, we demonstrate the successful synthesis, integration, and characterization of hybrid 2D material systems consisting of tandem repeat (TR) proteins inspired by squid ring teeth and graphene oxide (GO). The TR protein layer acts as a unique programmable molecular spacer between GO layers. As an application, we further demonstrate thermal actuation using bimorph molecular composite films. Bimorph actuators made of molecular composite films (GO/TR) can lead to energy efficiencies 18 times higher than regular bimorph actuators consisting of a GO layer and a TR protein layer (i.e., conventional bulk composite of GO and TR). Additionally, molecular composite bimorph actuators can reach curvature values as high as 1.2 cm-1 by using TR proteins with higher molecular weight, which is 3 times higher than conventional GO and TR composites.

cond-mat.soft↗

arXiv2011

Evaluation of the Neutron Background in a Direct WIMP Detector with Neutron Veto System Based on Gd-doped Liquid Scintillator

A direct WIMP (Weakly Interacting Massive Particle) detector with a neutron veto system is designed to better reject neutrons. Two experimental configurations are studied in the present paper: one is for an Xe detector enclosed in a Gd-loaded scintillator and the other one is for an Xe detector placed inside a reactor neutrino detector. The Gd-doped liquid scintillator (or the neutrino detector) is used as a neutron veto device. The neutron backgrounds for the two experimental designs have been estimated using Geant4 simulations. The results show that the neutron backgrounds can decrease to O(0.1) events per year per tonne of liquid Xenon. We calculate the sensitivities to spin-independent WIMP-nucleon elastic scattering. An exposure of one tonne $\times$ year could reach a cross-section of about 6$\times$$10^{-11}$ pb.

astro-ph.IM↗