arXiv subjects

Tao Xiang

Publications and source records attributed to Tao Xiang.

At least 361 records · Page 20Linked to original sources

arXiv2015

Semantic Graph for Zero-Shot Learning

Zero-shot learning aims to classify visual objects without any training data via knowledge transfer between seen and unseen classes. This is typically achieved by exploring a semantic embedding space where the seen and unseen classes can be related. Previous works differ in what embedding space is used and how different classes and a test image can be related. In this paper, we utilize the annotation-free semantic word space for the former and focus on solving the latter issue of modeling relatedness. Specifically, in contrast to previous work which ignores the semantic relationships between seen classes and focus merely on those between seen and unseen classes, in this paper a novel approach based on a semantic graph is proposed to represent the relationships between all the seen and unseen class in a semantic word space. Based on this semantic graph, we design a special absorbing Markov chain process, in which each unseen class is viewed as an absorbing state. After incorporating one test image into the semantic graph, the absorbing probabilities from the test data to each unseen class can be effectively computed; and zero-shot classification can be achieved by finding the class label with the highest absorbing probability. The proposed model has a closed-form solution which is linear with respect to the number of test images. We demonstrate the effectiveness and computational efficiency of the proposed method over the state-of-the-arts on the AwA (animals with attributes) dataset.

cs.CV↗

arXiv2015

Weakly Supervised Learning of Objects, Attributes and their Associations

When humans describe images they tend to use combinations of nouns and adjectives, corresponding to objects and their associated attributes respectively. To generate such a description automatically, one needs to model objects, attributes and their associations. Conventional methods require strong annotation of object and attribute locations, making them less scalable. In this paper, we model object-attribute associations from weakly labelled images, such as those widely available on media sharing sites (e.g. Flickr), where only image-level labels (either object or attributes) are given, without their locations and associations. This is achieved by introducing a novel weakly supervised non-parametric Bayesian model. Once learned, given a new image, our model can describe the image, including objects, attributes and their associations, as well as their locations and segmentation. Extensive experiments on benchmark datasets demonstrate that our weakly supervised model performs at par with strongly supervised models on tasks such as image description and retrieval based on object-attribute associations.

cs.CV↗

arXiv2015

Transductive Multi-label Zero-shot Learning

Zero-shot learning has received increasing interest as a means to alleviate the often prohibitive expense of annotating training data for large scale recognition problems. These methods have achieved great success via learning intermediate semantic representations in the form of attributes and more recently, semantic word vectors. However, they have thus far been constrained to the single-label case, in contrast to the growing popularity and importance of more realistic multi-label data. In this paper, for the first time, we investigate and formalise a general framework for multi-label zero-shot learning, addressing the unique challenge therein: how to exploit multi-label correlation at test time with no training data for those classes? In particular, we propose (1) a multi-output deep regression model to project an image into a semantic word space, which explicitly exploits the correlations in the intermediate semantic layer of word vectors; (2) a novel zero-shot learning algorithm for multi-label data that exploits the unique compositionality property of semantic word vector representations; and (3) a transductive learning strategy to enable the regression model learned from seen classes to generalise well to unseen classes. Our zero-shot learning experiments on a number of standard multi-label datasets demonstrate that our method outperforms a variety of baselines.

cs.LG↗

arXiv2015

Transductive Multi-class and Multi-label Zero-shot Learning

Recently, zero-shot learning (ZSL) has received increasing interest. The key idea underpinning existing ZSL approaches is to exploit knowledge transfer via an intermediate-level semantic representation which is assumed to be shared between the auxiliary and target datasets, and is used to bridge between these domains for knowledge transfer. The semantic representation used in existing approaches varies from visual attributes to semantic word vectors and semantic relatedness. However, the overall pipeline is similar: a projection mapping low-level features to the semantic representation is learned from the auxiliary dataset by either classification or regression models and applied directly to map each instance into the same semantic representation space where a zero-shot classifier is used to recognise the unseen target class instances with a single known 'prototype' of each target class. In this paper we discuss two related lines of work improving the conventional approach: exploiting transductive learning ZSL, and generalising ZSL to the multi-label case.

cs.LG↗

arXiv2015

Transductive Multi-view Zero-Shot Learning

Most existing zero-shot learning approaches exploit transfer learning via an intermediate-level semantic representation shared between an annotated auxiliary dataset and a target dataset with different classes and no annotation. A projection from a low-level feature space to the semantic representation space is learned from the auxiliary dataset and is applied without adaptation to the target dataset. In this paper we identify two inherent limitations with these approaches. First, due to having disjoint and potentially unrelated classes, the projection functions learned from the auxiliary dataset/domain are biased when applied directly to the target dataset/domain. We call this problem the projection domain shift problem and propose a novel framework, transductive multi-view embedding, to solve it. The second limitation is the prototype sparsity problem which refers to the fact that for each target class, only a single prototype is available for zero-shot learning given a semantic representation. To overcome this problem, a novel heterogeneous multi-view hypergraph label propagation method is formulated for zero-shot learning in the transductive embedding space. It effectively exploits the complementary information offered by different semantic representations and takes advantage of the manifold structures of multiple representation spaces in a coherent manner. We demonstrate through extensive experiments that the proposed approach (1) rectifies the projection shift between the auxiliary and target domains, (2) exploits the complementarity of multiple semantic representations, (3) significantly outperforms existing methods for both zero-shot and N-shot recognition on three image and video benchmark datasets, and (4) enables novel cross-view annotation tasks.

cs.CV↗

arXiv2014

Can Image-Level Labels Replace Pixel-Level Labels for Image Parsing

This paper presents a weakly supervised sparse learning approach to the problem of noisily tagged image parsing, or segmenting all the objects within a noisily tagged image and identifying their categories (i.e. tags). Different from the traditional image parsing that takes pixel-level labels as strong supervisory information, our noisily tagged image parsing is provided with noisy tags of all the images (i.e. image-level labels), which is a natural setting for social image collections (e.g. Flickr). By oversegmenting all the images into regions, we formulate noisily tagged image parsing as a weakly supervised sparse learning problem over all the regions, where the initial labels of each region are inferred from image-level labels. Furthermore, we develop an efficient algorithm to solve such weakly supervised sparse learning problem. The experimental results on two benchmark datasets show the effectiveness of our approach. More notably, the reported surprising results shed some light on answering the question: can image-level labels replace pixel-level labels (hard to access) as supervisory information for image parsing.

cs.CV↗

arXiv2014

Iron based high transition temperature superconductors

In a superconductor electrons form pairs and electric transport becomes dissipation-less at low temperatures. Recently discovered iron based superconductors have the highest superconducting transition temperature next to copper oxides. In this article, we review material aspects and physical properties of iron based superconductors. We discuss the dependence of transition temperature on the crystal structure, the interplay between antiferromagnetism and superconductivity by examining neutron scattering experiments, and the electronic properties of these compounds obtained by angle resolved photoemission spectroscopy in link with some results from scanning tunneling microscopy/spectroscopy measurements. Possible microscopic model for this class of compounds is discussed from a strong coupling point of view.

cond-mat.mtrl-sci↗

arXiv2014

Long-time dynamics of quantum chains: transfer-matrix renormalization group and entanglement of the maximal eigenvector

By using a different quantum-to-classical mapping from the Trotter-Suzuki decomposition, we identify the entanglement structure of the maximal eigenvectors for the associated quantum transfer matrix. This observation provides a deeper insight into the problem of linear growth of the entanglement entropy in time evolution using conventional methods. Based on this observation, we propose a general method for arbitrary temperatures using the biorthonormal transfer-matrix renormalization group. Our method exhibits a competitive accuracy with a much cheaper computational cost in comparison with two recent proposed methods for long-time dynamics based on a folding algorithm [Phys. Rev. Lett. 102, 240603 (2009)] and a modified time-dependent density-matrix renormalization group [Phys. Rev. Lett. 108, 227206 (2012)].

cond-mat.str-el↗

arXiv2013

Can deeply underdoped superconducting cuprates be topological superconductors?

The nodal $d_{x^2-y^2}$ superconducting gap is a hallmark of the cuprate high T$_c$ superconductors. Surprisingly recent angle-resolved photoemission spectroscopy of deeply underdoped cuprates revealed a nodeless energy gap which is adhered to the Fermi surface. Importantly this phenomenon is observed for compounds across several different cuprate families. In this letter we propose an exciting possibility, namely the fully gapped state is a topological superconductor.

cond-mat.supr-con↗

arXiv2013

Comparing Tensor Renormalization Group and Monte Carlo calculations for spin and gauge models

We show that the Tensor Renormalization Group (TRG) method can be applied to O(N) spin models, principal chiral models and pure gauge theories (Z2, U(1) and SU(2)) on (hyper) cubic lattices. We explain that contrarily to some common belief, it is very difficult to write compact formulas expressing the blockspinning of lattice models. We show that in contrast to other approaches, the TRG formulation allows us to write exact blocking formulas with numerically controllable truncations. The basic reason is that the TRG blocking separates neatly the degrees of freedom inside the block and which are integrated over, from those kept to communicate with the neighboring blocks. We argue that the TRG is a method that can handle large volumes, which is crucial to approach quasi-conformal systems. The method can also get rid of some sign problems. We discuss recent results regarding the critical properties of the 2D O(2) nonlinear sigma model with complex beta and chemical potential. As some of these results appeared in a recently published paper (PRD 88, 056005) and two recent preprints (arXiv:1309.4963 and arXiv:1309.6623), these proceedings rather emphasize the conceptual aspects of our ongoing effort.

hep-lat↗

arXiv2013

In-plane spin excitation anisotropy in the paramagnetic phase of NaFeAs

We use unpolarized and polarized inelastic neutron scattering to study low-energy spin excitations in NaFeAs, which exhibits a tetragonal-to-orthorhombic lattice distortion at $T_s\approx 58$ K followed by a collinear antiferromagnetic (AF) order below $T_N\approx 45$ K. In the AF ordered state ($T T_s$), spin excitations become more isotropic. Since similar magnetic anisotropy is also observed in the paramagnetic tetragonal phase of superconducting BaFe$_{1.904}$Ni$_{0.096}$As$_2$, our results suggest that the spin excitation anisotropy in superconducting iron pnictides originates from similar anisotropy already present in their parent compounds.

cond-mat.supr-con↗

arXiv2013

Investigation of atomic and electronic structures of MgOFeSe studied by the first-principles calculations

In order to assist the search of new superconductors in iron selenide materials by intercalation, we calculate the crystal and electronic structures of MgOFeSe using the first-principles density functional theory. MgOFeSe is isotructural to the parent compound of iron pnictide superconductor LaOFeAs. In LaOFeAs, the anion O$^{2-}$ is located at the center of each LaO tetrahedra. But for MgOFeSe, we find that the crystal structure with the cation Mg$^{2+}$ as the tetrahedral center in the MgO layer is energetically more stable. The low energy band structures around the Fermi surfaces of MgOFeSe are contributed mainly by Fe 3$d$ orbitals. The ground state of MgOFeSe is collinearly antiferromagnetically ordered. The height of Se atoms above the Fe-Fe layer is about 1.38 Å, which is close to the height of As from the Fe-Fe layer in the iron pnictide superconductors with optimal superconducting transition temperatures.

cond-mat.supr-con↗

arXiv2013

Prediction of phonon-mediated high temperature superconductivity in stoichiometric Li$_2$B$_3$C

The discovery of superconductivity in Magnesium Diborate (MgB$_2$) has stimulated great interest in the search of new superconductors with similar lattice structures. Unlike cuprate or iron-based superconductors, MgB$_2$ is indisputably a phonon-mediated high temperature superconductor. The emergence of high temperature superconductivity in this material results from the strong coupling between the boron $σ$-bonding electrons around the Fermi level and the bond-stretching optical phonon modes. Here we show, based on the first-principles calculations, that Li$_2$B$_3$C is such a good candidate of superconductor whose superconducting transition temperature (T$_c$) might be even higher than MgB$_2$. Li$_2$B$_3$C consists of alternating graphene-like boron-carbon layers and boron-boron layers with intercalated lithium atoms between them. Similar to MgB$_2$, Li$_2$B$_3$C is inherently metallic and possesses two $σ$- and two $π$-electron bands around the Fermi energy. The superconducting pairs are glued predominately by the strong interaction between boron $σ$-bonding electrons and various optical phonon modes.

cond-mat.supr-con↗

arXiv2012

Efficient simulation of infinite tree tensor network states on the Bethe lattice

We show that the simple update approach proposed by Jiang et. al. [H.C. Jiang, Z.Y. Weng, and T. Xiang, Phys. Rev. Lett. 101, 090603 (2008)] is an efficient and accurate method for determining the infinite tree tensor network states on the Bethe lattice. Ground state properties of the quantum transverse Ising model and the Heisenberg XXZ model on the Bethe lattice are studied. The transverse Ising model is found to undergo a second-order quantum phase transition with a diverging magnetic susceptibility but a finite correlation length which is upper-bounded by 1/ln(q-1) even at the transition point (q is the coordinate number of the Bethe lattice). An intuitive explanation on this peculiar "critical" phenomenon is given. The XXZ model on the Bethe lattice undergoes a first-order quantum phase transition at the isotropic point. Furthermore, the simple update scheme is found to be related with the Bethe approximation. Finally, by applying the simple update to various tree tensor clusters, we can obtain rather nice and scalable approximations for two-dimensional lattices.

cond-mat.str-el↗

arXiv2012

Effect of Li-deficiency impurities on the electron-overdoped LiFeAs superconductor

We use transport, inelastic neutron scattering, and angle resolved photoemission experiments to demonstrate that the stoichiometric LiFeAs is an intrinsically electron-overdoped superconductor similar to those of the electron-overdoped NaFe1-xTxAs and BaFe2-xTxAs2 (T = Co,Ni). Furthermore, we show that although transport properties of the stoichiometric superconducting LiFeAs and Li-deficient nonsuperconducting Li1-xFeAs are different, their electronic and magnetic properties are rather similar. Therefore, the nonsuperconducting Li1-xFeAs is also in the electron overdoped regime, where small Li deficiencies near the FeAs octahedra can dramatically suppress superconductivity through the impurity scattering effect.

cond-mat.supr-con↗

arXiv2012

Proximity effect at superconducting Sn-Bi2Se3 interface

We have investigated the conductance spectra of Sn-Bi2Se3 interface junctions down to 250 mK and in different magnetic fields. A number of conductance anomalies were observed below the superconducting transition temperature of Sn, including a small gap different from that of Sn, and a zero-bias conductance peak growing up at lower temperatures. We discussed the possible origins of the smaller gap and the zero-bias conductance peak. These phenomena support that a proximity-effect-induced chiral superconducting phase is formed at the interface between the superconducting Sn and the strong spin-orbit coupling material Bi2Se3.

cond-mat.supr-con↗

arXiv2012

Atomic and electronic structures of FeSe monolayer and bilayer thin films on SrTiO$_3$ (001): a first-principles study

By the first-principles electronic structure calculations, we have studied electronic structures of FeSe monolayer and bilayer thin films on SrTiO$_3$ (001) with SrO-termination or TiO$_2$ termination. We find that both FeSe monolayer and bilayer on either termination behave like a slightly doped semiconductor and a collinear antiferromagnetic order on Fe ions. There is no substantial charge transfer between the FeSe layers and the substrate. FeSe is adhered to the SrTiO$_3$ surface by a dipole-dipole interaction. The Fermi surface is mainly the contribution of Fe-3d orbitals. A valence band contributed mainly by the O-$2p$ orbitals in the TiO$_2$ layer is located slightly below the Fermi level, which can become conducting upon a small doping of holes.

cond-mat.supr-con↗

arXiv2012

(PI,0) antiferromagnetic spin excitations in superconducting Rb0.82Fe1.68Se2

We use inelastic neutron scattering to show that superconducting (SC) rubidium iron selenide Rb0.82Fe1.68Se2 exhibits antiferromagnetic (AF) spin excitations near the in-plane wave vector Q = (PI, 0) identical to that for iron arsenide superconductors. Moreover, we find that these excitations change from incommensurate to commensurate with increasing energy, and occur at the expense of spin waves associated with the coexisting sqrt(5)\timessqrt(5) block AF phase. Since angle resolved photoemission experiments reveal no evidence for hole-like Fermi surface at Gamma(0, 0), our results suggest that the Q = (PI, 0) excitations in SC Rb0.82Fe1.68Se2 come from localized moments and may have a similar origin as the hourglass-like spin excitations in copper oxide superconductors.

cond-mat.supr-con↗