arXiv subjects

Mao Ye

Publications and source records attributed to Mao Ye.

144 records · Page 8Linked to original sources

arXiv2019

Distribution-Aware Coordinate Representation for Human Pose Estimation

While being the de facto standard coordinate representation in human pose estimation, heatmap is never systematically investigated in the literature, to our best knowledge. This work fills this gap by studying the coordinate representation with a particular focus on the heatmap. Interestingly, we found that the process of decoding the predicted heatmaps into the final joint coordinates in the original image space is surprisingly significant for human pose estimation performance, which nevertheless was not recognised before. In light of the discovered importance, we further probe the design limitations of the standard coordinate decoding method widely used by existing methods, and propose a more principled distribution-aware decoding method. Meanwhile, we improve the standard coordinate encoding process (i.e. transforming ground-truth coordinates to heatmaps) by generating accurate heatmap distributions for unbiased model training. Taking the two together, we formulate a novel Distribution-Aware coordinate Representation of Keypoint (DARK) method. Serving as a model-agnostic plug-in, DARK significantly improves the performance of a variety of state-of-the-art human pose estimation models. Extensive experiments show that DARK yields the best results on two common benchmarks, MPII and COCO, consistently validating the usefulness and effectiveness of our novel coordinate representation idea.

cs.CV↗

arXiv2019

Observation of a Room Temperature Two-dimensional Ferroelectric Metal

Materials with reduced dimensions have been shown to host a wide variety of exotic properties and novel quantum states that often defy textbook wisdom1-5. Ferroelectric polarization and metallicity are well-known examples of mutually exclusive properties that cannot coexist in bulk solids because the net electric field in a metal can be fully screened by free electrons6. An atomically thin metallic layer capped by insulating layers has shown decent conductivity at room temperature7. Moreover, a penetrating polarization field can be employed to induce an ion displacement and create an intrinsic polarization in the metallic layer. Here we demonstrate that a ferroelectric metal can be artificially synthesized through imposing a strong polarization field in the form of ferroelectric/unit-cell-thin metal superlattices. In this way the symmetry of an atomically thin conductive layer can be broken and manipulated by a neighboring polar field, thereby forming a two-dimensional (2D) ferroelectric metal. The fabricated of (SrRuO3)1/(BaTiO3)10 superlattices exhibit ferroelectric polarization in an atomically thin layer with metallic conductivity at room temperature. A multipronged investigation combining structural analyses, electrical measurements, and first-principles electronic structure calculations unravels the coexistence of 2D electrical conductivity in the SrRuO3 monolayer accompanied by the electric polarization. Such 2D ferroelectric metal paves a novel way to engineer a quantum multi-state with unusual coexisting properties, such as ferroelectrics, ferromagnetics and metals, manipulated by external fields8,9.

cond-mat.mtrl-sci↗

arXiv2019

Metalens With Artificial Focus Pattern

Metalens as one of the most popular applications of emmerging optical metasurfaces has raised widspread interest recently. With nano structures fully controlling phase, polarization and transmission, metalens has achieved comparable performance of commercial objective lenses. While recent studies seeking for the accomplishment of traditional focusing behaviors through metalens are successful, inthis work, we have discovered that instead of focusing light to a point, metasurface further enables shaping the focus into a flexibly designed pattern, with more promises and potentials. New mechanism and generalizations of conventional point-focused metalens guiding principles have been proposed with metalens concentrating light to artificial focus pattern. As proving examples, we have demonstrated the engineering of metalens with artificial focus pattern by creating line and ring-shaped focus as 'drawing tools'. The metalens with 'U' and 'M' shaped focus are characterized for the proof of concepts. These metalens are fabricated through a single layer of silicon-based material through CMOS compatible nano fabrication process. The mechanism to generate artificial focus pattern can be applied to a plethora of future on-chip optical devices with applications ranging from beam engineering to next generation nano lithography.

eess.IV↗

arXiv2019

Fast Human Pose Estimation

Existing human pose estimation approaches often only consider how to improve the model generalisation performance, but putting aside the significant efficiency problem. This leads to the development of heavy models with poor scalability and cost-effectiveness in practical use. In this work, we investigate the under-studied but practically critical pose model efficiency problem. To this end, we present a new Fast Pose Distillation (FPD) model learning strategy. Specifically, the FPD trains a lightweight pose neural network architecture capable of executing rapidly with low computational cost. It is achieved by effectively transferring the pose structure knowledge of a strong teacher network. Extensive evaluations demonstrate the advantages of our FPD method over a broad range of state-of-the-art pose estimation approaches in terms of model cost-effectiveness on two standard benchmark datasets, MPII Human Pose and Leeds Sports Pose.

cs.CV↗

arXiv2018

Triad-NVM: Persistent-Security for Integrity-Protected and Encrypted Non-Volatile Memories (NVMs)

Emerging Non-Volatile Memories (NVMs) are promising contenders for building future memory systems. On the other side, unlike DRAM systems, NVMs can retain data even after power loss and thus enlarge the attack surface. While data encryption and integrity verification have been proposed earlier for DRAM systems, protecting and recovering secure memories becomes more challenging with persistent memory. Specifically, security metadata, e.g., encryption counters and Merkle Tree data, should be securely persisted and recovered across system reboots and during recovery from crashes. Not persisting updates to security metadata can lead to data inconsistency, in addition to serious security vulnerabilities. In this paper, we pioneer a new direction that explores persistency of both Merkle Tree and encryption counters to enable secure recovery of data-verifiable and encrypted memory systems. To this end, we coin a new concept that we call Persistent-Security. We discuss the requirements for such persistently secure systems, propose novel optimizations, and evaluate the impact of the proposed relaxation schemes and optimizations on performance, resilience and recovery time. To the best of our knowledge, our paper is the first to discuss the persistence of security metadata in integrity-protected NVM systems and provide corresponding optimizations. We define a set of relaxation schemes that bring trade-offs between performance and recovery time for large capacity NVM systems. Our results show that our proposed design, Triad-NVM, can improve the throughput by an average of ~2x (relative to strict persistence). Moreover, Triad-NVM maintains a recovery time of less than 4 seconds for an 8TB NVM system (30.6 seconds for 64TB), which is ~3648x faster than a system without security metadata persistence.

cs.DC↗

arXiv2018

Possible structural origin of superconductivity in Sr-doped Bi2Se3

Doping bismuth selenide (Bi2Se3) with elements such as copper and strontium (Sr) can induce superconductivity, making the doped materials interesting candidates to explore potential topological superconducting behaviors. It was thought that the superconductivity of doped Bi2Se3 was induced by dopant atoms intercalated in van der Waals gaps. However, several experiments have shown that the intercalation of dopant atoms may not necessarily make doped Bi2Se3 superconducting. Thus, the structural origin of superconductivity in doped Bi2Se3 remains an open question. Herein, we combined material synthesis and characterization, high-resolution transmission electron microscopy, and first-principles calculations to study the doping structure of Sr-doped Bi2Se3. We found that the emergence of superconductivity is strongly related with n-type dopant atoms. Atomic-level energy-dispersive X-ray mapping revealed various n-type Sr dopants that occupy intercalated and interstitial positions. First-principles calculations showed that the formation energy of a specific interstitial Sr doping position depends strongly on Sr doping level. This site changes from a metastable position at low Sr doping level to a stable position at high Sr doping level. The calculation results explain why quenching is necessary to obtain superconducting samples when the Sr doping level is low and also why slow furnace cooling can yield superconducting samples when the Sr doping level is high. Our findings suggest that Sr atoms doped at interstitial locations, instead of those intercalated in van der Waals gaps, are most likely to be responsible for the emergence of superconductivity in Sr-doped Bi2Se3.

cond-mat.supr-con↗

arXiv2017

Do Convolutional Neural Networks Learn Class Hierarchy?

Convolutional Neural Networks (CNNs) currently achieve state-of-the-art accuracy in image classification. With a growing number of classes, the accuracy usually drops as the possibilities of confusion increase. Interestingly, the class confusion patterns follow a hierarchical structure over the classes. We present visual-analytics methods to reveal and analyze this hierarchy of similar classes in relation with CNN-internal data. We found that this hierarchy not only dictates the confusion patterns between the classes, it furthermore dictates the learning behavior of CNNs. In particular, the early layers in these networks develop feature detectors that can separate high-level groups of classes quite well, even after a few training epochs. In contrast, the latter layers require substantially more epochs to develop specialized feature detectors that can separate individual classes. We demonstrate how these insights are key to significant improvement in accuracy by designing hierarchy-aware CNNs that accelerate model convergence and alleviate overfitting. We further demonstrate how our methods help in identifying various quality issues in the training data.

cs.CV↗

arXiv2017

Pose-Invariant Face Alignment with a Single CNN

Face alignment has witnessed substantial progress in the last decade. One of the recent focuses has been aligning a dense 3D face shape to face images with large head poses. The dominant technology used is based on the cascade of regressors, e.g., CNN, which has shown promising results. Nonetheless, the cascade of CNNs suffers from several drawbacks, e.g., lack of end-to-end training, hand-crafted features and slow training speed. To address these issues, we propose a new layer, named visualization layer, that can be integrated into the CNN architecture and enables joint optimization with different loss functions. Extensive evaluation of the proposed method on multiple datasets demonstrates state-of-the-art accuracy, while reducing the training time by more than half compared to the typical cascade of CNNs. In addition, we compare multiple CNN architectures with the visualization layer to further demonstrate the advantage of its utilization.

cs.CV↗

arXiv2015

Ultrafast electron dynamics at the Dirac node of the topological insulator Sb$_2$Te$_3$

Topological insulators (TIs) are a new quantum state of matter. Their surfaces and interfaces act as a topological boundary to generate massless Dirac fermions with spin-helical textures. Investigation of fermion dynamics near the Dirac point is crucial for the future development of spintronic devices incorporating topological insulators. However, research so far has been unsatisfactory because of a substantial overlap with the bulk valence band and a lack of a completely unoccupied Dirac point (DP). Here, we explore the surface Dirac fermion dynamics in the TI Sb$_2$Te$_3$ by time- and angle-resolved photoemission spectroscopy (TrARPES). Sb$_2$Te$_3$ has a DP located completely above the Fermi energy ($E_F$) with an in-gap DP. The excited electrons in the upper Dirac cone stay longer than those below the Dirac point to form an inverted population. This was attributed to a reduced density of states (DOS) near the DP .

cond-mat.mtrl-sci↗

arXiv2015

Multi-channel exchange-scattering spin polarimetry

Electron spin takes critical role in almost all novel phenomena discovered in modern condensed matter physics (High-temperature superconductivity, Kondo effect, Giant Magnetoresistance, topological insulator, quantum anomalous Hall effect, etc.). However, the measurements for electron spin is of poor quality which blocks the development of material sciences because of the low efficiency of spin polarimeter. Here we show an imaging type exchange-scattering spin polarimeter with 5 orders more efficiency compared with a classical Mott polarimeter. As a demonstration, the fine spin structure of electronic states in bismuth (111) is investigated, showing the strong Rashba type spin splitting behavior in both bulk and surface states. This improvement pave the way to study novel spin related phenomena with unprecedented accuracy.

cond-mat.mtrl-sci↗

arXiv2014

The design of a high flux VUV beamline for low energy photons

A VUV beamline at SSRF for ARPES measurements are designed. To increase the resolution and bulk sensitivity, the photon energy as low as 7 eV is desired. Because the reflectivity for p-polarized photons strongly decreases when the photon energy is below 30 eV, the design of high flux beamline for low energy VUV photons is a challenge. This work shows a variable including angle VLPGM with varied grating depth (VGD) which can achieve both high resolution and high flux with broad energy coverage.

physics.ins-det↗

arXiv2014

The design of a multi-channel spin polarimeter

All commercial electron spin polarimeters work in single channel mode, which is the bottleneck of researches by spin-resolved photoelectron spectroscopy. By adopting the time inversion antisymmetry of magnetic field, we developed a multichannel spin polarimeter based on normal incident VLEED. The key point to achieve the multi-channel measurements is the spatial resolution of the electron optics. The test of the electron optics shows that the designed spatial resolution can be achieved and an image type spin polarimeter with 100 times 100, totally ten thousand channels is possible to be realized.

physics.ins-det↗

arXiv2014

Realization of bulk insulating property and carrier manipulation in reversible spin current regime of the ideal topological insulator TlBiSe2

The surfaces of three-dimensional topological insulators (TIs) characterized by a spin-helical Dirac fermion provide a fertile ground for realizing exotic phenomena as well as having potential for wide-ranging applications. To realize most of their special properties, the Dirac point (DP) is required to be located near the Fermi energy with a bulk insulating property while it is hardly achieved in most of the discovered TIs. It has been recently found that TlBiSe2 features an in-gap DP, where upper and lower parts of surface Dirac cone are both utilized. Nevertheless, investigations of the surface transport properties of this material are limited due to the lack of bulk insulating characteristics. Here, we present the first realization of bulk insulating property by tuning the composition of Tl1-xBi1+xSe2-d without introducing guest atoms that can bring the novel properties into the reality. This result promises to shed light on new exotic topological phenomena on the surface.

cond-mat.mes-hall↗

arXiv2013

Experimental verification of the surface termination in the topological insulator TlBiSe$_{2}$ using core-level photoelectron spectroscopy and scanning tunneling microscopy

The surface termination of the promising topological insulator TlBiSe$_{2}$ has been studied by surface and bulk sensitive probes. Our scanning tunneling microscopy has unmasked for the first time the unusual surface morphology of TlBiSe$_{2}$ obtained by cleaving, where islands are formed by residual atoms on the cleaved plane. The chemical condition of these islands was identified using core-level spectroscopy. We observed thallium core-level spectra that are strongly deformed by a surface component in sharp contrast to the other elements. We propose a simple explanation for this behavior by assuming that the sample cleaving breaks the bonding between thallium and selenium atoms, leaving the thallium layer partially covering the selenium layer. These findings will assist the interpretation of future experimental and theoretical studies on this surface.

cond-mat.mes-hall↗

arXiv2012

From User Comments to On-line Conversations

We present an analysis of user conversations in on-line social media and their evolution over time. We propose a dynamic model that accurately predicts the growth dynamics and structural properties of conversation threads. The model successfully reconciles the differing observations that have been reported in existing studies. By separating artificial factors from user behaviors, we show that there are actually underlying rules in common for on-line conversations in different social media websites. Results of our model are supported by empirical measurements throughout a number of different social media websites.

cs.CY↗

arXiv2011

Collective Attention and the Dynamics of Group Deals

We present a study of the group purchasing behavior of daily deals in Groupon and LivingSocial and introduce a predictive dynamic model of collective attention for group buying behavior. In our model, the aggregate number of purchases at a given time comprises two types of processes: random discovery and social propagation. We find that these processes are very clearly separated by an inflection point. Using large data sets from both Groupon and LivingSocial we show how the model is able to predict the success of group deals as a function of time. We find that Groupon deals are easier to predict accurately earlier in the deal lifecycle than LivingSocial deals due to the final number of deal purchases saturating quicker. One possible explanation for this is that the incentive to socially propagate a deal is based on an individual threshold in LivingSocial, whereas in Groupon it is based on a collective threshold, which is reached very early. Furthermore, the personal benefit of propagating a deal is also greater in LivingSocial.

cs.CY↗

arXiv2011

Exploring Social Influence for Recommendation - A Probabilistic Generative Model Approach

In this paper, we propose a probabilistic generative model, called unified model, which naturally unifies the ideas of social influence, collaborative filtering and content-based methods for item recommendation. To address the issue of hidden social influence, we devise new algorithms to learn the model parameters of our proposal based on expectation maximization (EM). In addition to a single-machine version of our EM algorithm, we further devise a parallelized implementation on the Map-Reduce framework to process two large-scale datasets we collect. Moreover, we show that the social influence obtained from our generative models can be used for group recommendation. Finally, we conduct comprehensive experiments using the datasets crawled from last.fm and whrrl.com to validate our ideas. Experimental results show that the generative models with social influence significantly outperform those without incorporating social influence. The unified generative model proposed in this paper obtains the best performance. Moreover, our study on social influence finds that users in whrrl.com are more likely to get influenced by friends than those in last.fm. The experimental results also confirm that our social influence based group recommendation algorithm outperforms the state-of-the-art algorithms for group recommendation.

cs.SI↗

arXiv2008

Co-induced nano-structures on Si(111) surface

The interaction of cobalt atoms with silicon (111) surface has been investigated by means of scanning tunneling microscopy (STM) and low-energy electron diffraction (LEED). Besides the Co silicide islands, we have successfully distinguished two inequivalent Co-induced $\sqrt{13}\times\sqrt{13}$ reconstructions on Si (111) surface. Our high-resolution STM images provide some structural properties of the two different $\sqrt{13}\times\sqrt{13}$ derived phases. Both of the two phases seem to form islands with single domain. The new findings will help us to understand the early stage of Co silicide formations.

cond-mat.mtrl-sci↗