arXiv subjects

Ye Yuan

Publications and source records attributed to Ye Yuan.

At least 343 records · Page 19Linked to original sources

arXiv2017

Video Representation Learning and Latent Concept Mining for Large-scale Multi-label Video Classification

We report on CMU Informedia Lab's system used in Google's YouTube 8 Million Video Understanding Challenge. In this multi-label video classification task, our pipeline achieved 84.675% and 84.662% GAP on our evaluation split and the official test set. We attribute the good performance to three components: 1) Refined video representation learning with residual links and hypercolumns 2) Latent concept mining which captures interactions among concepts. 3) Learning with temporal segments and weighted multi-model ensemble. We conduct experiments to validate and analyze the contribution of our models. We also share some unsuccessful trials leveraging conventional approaches such as recurrent neural networks for video representation learning for this large-scale video dataset. All the codes to reproduce our results are publicly available at https://github.com/Martini09/informedia-yt8m-release.

cs.CV↗

arXiv2017

Interaction between magnetic moments and itinerant carriers in d0 ferromagnetic SiC

Elucidating the interaction between magnetic moments and itinerant carriers is an important step to spintronic applications. Here, we investigate magnetic and transport properties in d0 ferromagnetic SiC single crystals prepared by postimplantation pulsed laser annealing. Magnetic moments are contributed by the p states of carbon atoms, but their magnetic circular dichroism is different from that in semi-insulating SiC samples. The anomalous Hall effect and negative magnetoresistance indicate the influence of d0 spin order on free carriers. The ferromagnetism is relatively weak in N-implanted SiC compared with that in Al-implanted SiC after annealing. The results suggest that d0 magnetic moments and itinerant carriers can interact with each other, which will facilitate the development of SiC spintronic devices with d0 ferromagnetism.

cond-mat.mtrl-sci↗

arXiv2017

Distributed Kalman filtering with minimum-time consensus algorithm

Fueled by applications in sensor networks, these years have witnessed a surge of interest in distributed estimation and filtering. A new approach is hereby proposed for the Distributed Kalman Filter (DKF) by integrating a local covariance computation scheme. Compared to existing well-established DKF methods, the virtue of the present approach lies in accelerating the convergence of the state estimates to those of the Centralized Kalman Filter (CKF). Meanwhile, an algorithm is proposed that allows each node to compute the averaged measurement noise covariance matrix within a minimal discrete-time running steps in a distributed way. Both theoretical analysis and extensive numerical simulations are conducted to show the feasibility and superiority of the proposed method.

eess.SY↗

arXiv2016

Edge pinning and transformation of defect lines induced by faceted colloidal rings in nematic liquid crystals

Nematic colloids exhibit a large diversity of topological defects and structures induced by colloidal particles in the orientationally ordered liquid crystal host fluids. These defects and field configurations define elastic interactions and medium-mediated self-assembly, as well as serve as model systems in exploiting the richness of interactions between topologies and geometries of colloidal surfaces, nematic fields, and topological singularities induced by particles in the nematic bulk and at nematic-colloidal interfaces. Here we demonstrate formation of quarter-strength surface-pinned disclinations, as well as a large variety of director field configurations with splitting and reconnections of singular defect lines, prompted by colloidal particles with sharp edges and size large enough to define strong boundary conditions. Using examples of faceted ring-shaped particles of genus g = 1, we explore transformation of defect lines as they migrate between locations in the bulk of the nematic host to edge-pinned locations at the surfaces of particles and vice versa, showing that this behavior is compliant with topological constraints defined by mathematical theorems. We discuss how transformation of bulk and surface defect lines induced by faceted colloids can enrich the diversity of elasticity-mediated colloidal interactions and how these findings may impinge on prospects of their controlled reconfigurable self-assembly in nematic hosts.

cond-mat.soft↗

arXiv2016

Joint Hand Detection and Rotation Estimation by Using CNN

Hand detection is essential for many hand related tasks, e.g. parsing hand pose, understanding gesture, which are extremely useful for robotics and human-computer interaction. However, hand detection in uncontrolled environments is challenging due to the flexibility of wrist joint and cluttered background. We propose a deep learning based approach which detects hands and calibrates in-plane rotation under supervision at the same time. To guarantee the recall, we propose a context aware proposal generation algorithm which significantly outperforms the selective search. We then design a convolutional neural network(CNN) which handles object rotation explicitly to jointly solve the object detection and rotation estimation tasks. Experiments show that our method achieves better results than state-of-the-art detection models on widely-used benchmarks such as Oxford and Egohands database. We further show that rotation estimation and classification can mutually benefit each other.

cs.CV↗

arXiv2016

Topological nanocolloids with facile electric switching of plasmonic properties

Combining topology and plasmonics paradigms in nanocolloidal systems may enable new means of pre-engineering desired composite material properties. Here we design and realize orientationally ordered assemblies of noble metal nanoparticles with genus-one topology and unusual long-range ordering mediated by their interactions with the surrounding nematic fluid host. Facile electric switching of these composites is reminiscent to that of pristine liquid crystals (LCs), but provides a means of reconfiguring the nanoparticle assembly and thus also the ensuing composite medium's optical properties. Our findings may lead to formation of new molecular-colloidal soft matter phases with unusual optical properties as well as optical metamaterials.

cond-mat.soft↗

arXiv2016

Event Detection and Localization in Distribution Grids with Phasor Measurement Units

The recent introduction of synchrophasor technology into power distribution systems has given impetus to various monitoring, diagnostic, and control applications, such as system identification and event detection, which are crucial for restoring service, preventing outages, and managing equipment health. Drawing on the existing framework for inferring topology and admittances of a power network from voltage and current phasor measurements, this paper proposes an online algorithm for event detection and localization in unbalanced three-phase distribution systems. Using a convex relaxation and a matrix partitioning technique, the proposed algorithm is capable of identifying topology changes and attributing them to specific categories of events. The performance of this algorithm is evaluated on a standard test distribution feeder with synthesized loads, and it is shown that a tripped line can be detected and localized in an accurate and timely fashion, highlighting its potential for real-world applications.

eess.SY↗

arXiv2016

Review Networks for Caption Generation

We propose a novel extension of the encoder-decoder framework, called a review network. The review network is generic and can enhance any existing encoder- decoder model: in this paper, we consider RNN decoders with both CNN and RNN encoders. The review network performs a number of review steps with attention mechanism on the encoder hidden states, and outputs a thought vector after each review step; the thought vectors are used as the input of the attention mechanism in the decoder. We show that conventional encoder-decoders are a special case of our framework. Empirically, we show that our framework improves over state-of- the-art encoder-decoder systems on the tasks of image captioning and source code captioning.

cs.LG↗

arXiv2016

Radiation Studies for the Target Station of the MOMENT

The discovery of the neutrino mixing angle $θ_{13}$ opens new opportunities for the discovery of the leptonic CP violation for high intensity neutrino beams. MOMENT a future neutrino facility with a high-power proton beam of 15 MW from a continuous-wave linac is focused on that discovery. The high power of the proton beam causes extreme radiation conditions for the facility and especially for the target station where the pion capture system of five superconducting solenoids is located. In this paper initial studies are performed for the effects of the radiation on the solenoid structure and the area surrounding it. A concept cooling system is also proposed.

physics.ins-det↗

arXiv2016

Novel dense crystalline packings of ellipsoids

An ellipsoid, the simplest non-spherical shape, has been extensively used as models for elongated building blocks for a wide spectrum of molecular, colloidal and granular systems. Yet the densest packing of congruent hard ellipsoids, which is intimately related to the high-density phase of many condensed matter systems, is still an open problem. We discover a novel dense crystalline packing of ellipsoids containing 24 particles with a quasi-square-triangular (SQ-TR) tiling arrangement, whose packing density ϕ exceeds that of the SM2 crystal for aspect ratios α in (1.365, 1.5625), attaining a maximal ϕ about 0.75806 at α = 93/64. We show that SQ-TR phase is thermodynamically stable at high densities over the aforementioned α range and report a novel phase diagram for self-dual ellipsoids. The discovery of SQ-TR crystal suggests novel organizing principles for non-spherical particles and self-assembly of colloidal systems.

cond-mat.stat-mech↗

arXiv2016

Suppressing the cellular breakdown in silicon supersaturated with titanium

Hyper doping Si with up to 6 at.% Ti in solid solution was performed by ion implantation followed by pulsed laser annealing and flash lamp annealing. In both cases, the implanted Si layer can be well recrystallized by liquid phase epitaxy and solid phase epitaxy, respectively. Cross-sectional transmission electron microscopy of Ti-implanted Si after liquid phase epitaxy shows the so-called growth interface breakdown or cellular breakdown owing to the occurrence of constitutional supercooling in the melt. The appearance of cellular breakdown prevents further recrystallization. However, the out-diffusion and cellular breakdown can be effectively suppressed by solid phase epitaxy during flash lamp annealing due to the high velocity of amorphous-crystalline interface and the low diffusion velocity for Ti in the solid phase.

cond-mat.mtrl-sci↗

arXiv2016

Precise tuning of the Curie temperature of (Ga,Mn)As-based magnetic semiconductors by hole compensation: Support for valence-band ferromagnetism

For the prototype diluted ferromagnetic semiconductor (Ga,Mn)As, there is a fundamental concern about the electronic states near the Fermi level, i.e., whether the Fermi level resides in a well-separated impurity band derived from Mn doping (impurity-band model) or in the valence band that is already merged with the Mn-derived impurity band (valence-band model). We investigate this question by carefully shifting the Fermi level by means of carrier compensation. We use helium-ion implantation, a standard industry technology, to precisely compensate the hole doping of GaAs-based diluted ferromagnetic semiconductors while keeping the Mn concentration constant. We monitor the change of Curie temperature ($T_C$) and conductivity. For a broad range of samples including (Ga,Mn)As and (Ga,Mn)(As,P) with various Mn and P concentrations, we observe a smooth decrease of $T_C$ with carrier compensation over a wide temperature range while the conduction is changed from metallic to insulating. The existence of $T_C$ below 10\,K is also confirmed in heavily compensated samples. Our experimental results are naturally explained within the valence-band picture.

cond-mat.mtrl-sci↗

arXiv2016

Properties of massive star-forming clumps with infall motions

In this work, we aim to characterise high-mass clumps with infall motions. We selected 327 clumps from the Millimetre Astronomy Legacy Team 90-GHz (MALT90) survey, and identified 100 infall candidates. Combined with the results of He et al. (2015), we obtained a sample of 732 high-mass clumps, including 231 massive infall candidates and 501 clumps where infall is not detected. Objects in our sample were classified as pre-stellar, proto-stellar, HII or photo-dissociation region (PDR). The detection rates of the infall candidates in the pre-stellar, proto-stellar, HII and PDR stages are 41.2%, 36.6%, 30.6% and 12.7%, respectively. The infall candidates have a higher H$_{2}$ column density and volume density compared with the clumps where infall is not detected at every stage. For the infall candidates, the median values of the infall rates at the pre-stellar, proto-stellar, HII and PDR stages are 2.6$\times$10$^{-3}$, 7.0$\times$10$^{-3}$, 6.5$\times$10$^{-3}$ and 5.5$\times$10$^{-3}$ M$_\odot$ yr$^{-1}$, respectively. These values indicate that infall candidates at later evolutionary stages are still accumulating material efficiently. It is interesting to find that both infall candidates and clumps where infall is not detected show a clear trend of increasing mass from the pre-stellar to proto-stellar, and to the HII stages. The power indices of the clump mass function (ClMF) are 2.04$\pm$0.16 and 2.17$\pm$0.31 for the infall candidates and clumps where infall is not detected, respectively, which agree well with the power index of the stellar initial mass function (2.35) and the cold Planck cores (2.0).

astro-ph.GA↗

arXiv2016

Track segment finding with CGEM-IT and matching to tracks in ODC

The relative differences in coordinates of Cylindrical-Gas-Electron-Multiplier-Detector-based Inner Tracker (CGEM-IT) clusters are studied to search for track segments in CGEM-IT. With the full simulation of single muon track samples, clear patterns are found and parameterized for the correct cluster combinations. The cluster combinations satisfying the patterns are selected as track segment candidates in CGEM-IT with an efficiency higher than 99%. The parameters of the track segments are obtained by a helix fitting. Some chi-squared quantities, evaluating the differences in track parameters between the track segments in CGEM-IT and the tracks found in Outer-Drift-Chamber (ODC), are calculated and used to match them. Proper chi-squared requirements are determined as a function of transverse momentum and the matching efficiency is found reasonable.

physics.ins-det↗

arXiv2016

Intrinsic diamagnetism in the Weyl semimetal TaAs

We investigate the magnetic properties of TaAs, a prototype Weyl semimetal. TaAs crystals show weak diamagnetism with magnetic susceptibility of about -7 * 10^{-7} emu/(g*Oe) at 5 K. A general feature is the appearance of a minimum at around 185 K in magnetization measurements as a function of temperature. No phase transition is observed in the temperature range between 5 K and 400 K. The magnetic properties indicate that the intrinsic Fermi level in TaAs is not located at the Weyl nodes, in agreement with the theory prediction.

cond-mat.mtrl-sci↗

arXiv2015

Study of the Tracking Method and Expected Performance of the Silicon Pixel Inner Tracker Applied in BESIII

The inner drift chamber of the BESIII is encountering serious aging problem after five year's running. For the first layer, the decrease in gas gain is about 26% from 2009 to 2013. The upgrade of the inner tracking detector has become an urgent problem for the BESIII experiment. An inner tracker using CMOS pixel sensors is an important candidate because of its great advantages on spatial resolution and radiation hardness. In order to carry out a Monte Carlo study on the expected performance, a Geant4-based full simulation for the silicon pixel detector has been implemented. The tracking method combining the silicon pixel inner tracker and outer drift chamber has been studied and a preliminary reconstruction software was developed. The Monte Carlo study shows that the performances including momentum resolution, vertex resolution and the tracking efficiency are significantly improved due to the good spatial resolution and moderate material budget of the silicon pixel detector.

physics.ins-det↗

arXiv2015

Study of tracking efficiency and its systematic uncertainty from $J/ψ\to p \overline{p} π^+ π^-$ at BESIII

Based on $J/ψ$ events collected with the BESIII detector, with corresponding Monte Carlo samples, the tracking efficiency and its systematic uncertainty are studied using a control sample of $J/ψ\to p \overline p π^+ π^-$. Validation methods and different factors influencing the tracking efficiency are presented in detail. The tracking efficiency and its systematic uncertainty for protons and pions with the transverse momentum and polar angle dependence are also discussed.

hep-ex↗

arXiv2015

Identifying Biochemical Reaction Networks From Heterogeneous Datasets

In this paper, we propose a new method to identify biochemical reaction networks (i.e. both reactions and kinetic parameters) from heterogeneous datasets. Such datasets can contain (a) data from several replicates of an experiment performed on a biological system; (b) data measured from a biochemical network subjected to different experimental conditions, for example, changes/perturbations in biological inductions, temperature, gene knock-out, gene over-expression, etc. Simultaneous integration of various datasets to perform system identification has the potential to avoid non-identifiability issues typically arising when only single datasets are used.

eess.SY↗