• ### Canonical Correlation Analysis of Datasets with a Common Source Graph(1803.10309)

March 27, 2018 cs.LG, stat.ML, cs.SY
Canonical correlation analysis (CCA) is a powerful technique for discovering whether or not hidden sources are commonly present in two (or more) datasets. Its well-appreciated merits include dimensionality reduction, clustering, classification, feature selection, and data fusion. The standard CCA however, does not exploit the geometry of the common sources, which may be available from the given data or can be deduced from (cross-) correlations. In this paper, this extra information provided by the common sources generating the data is encoded in a graph, and is invoked as a graph regularizer. This leads to a novel graph-regularized CCA approach, that is termed graph (g) CCA. The novel gCCA accounts for the graph-induced knowledge of common sources, while minimizing the distance between the wanted canonical variables. Tailored for diverse practical settings where the number of data is smaller than the data vector dimensions, the dual formulation of gCCA is also developed. One such setting includes kernels that are incorporated to account for nonlinear data dependencies. The resultant graph-kernel (gk) CCA is also obtained in closed form. Finally, corroborating image classification tests over several real datasets are presented to showcase the merits of the novel linear, dual, and kernel approaches relative to competing alternatives.
• ### Tunable Electronic Structure and Topological Properties of $LnPn$ ($Ln$=Ce, Pr, Gd, Sm, Yb; $Pn$=Sb, Bi)(1802.04554)

We have performed systematic first principles study of the electronic structure and band topology properties of $LnPn$ compounds ($Ln$=Ce, Pr, Gd, Sm, Yb; $Pn$=Sb, Bi). Assuming the $f$-electrons are well localized in these materials, both hybrid functional and modified Becke-Johnson calculations yield electronic structure in good agreement with experimental observations, while generalized gradient approximation calculations severely overestimate the band inversions. From Ce to Yb, a systematic reduction of band inversion with respect to the increasing $Ln$ atomic number is observed, and $\mathcal{Z}_2$ for Ce$Pn$ and Yb$Pn$ are [1;000] and [0;000], respectively. In both hybrid functional and modified Becke-Johns calculations, a topologically nontrivial to trivial transition is expected around SmSb for the antimonides and around DyBi for the bismuthides. Such variation is related with lanthanide contraction, but is different from simple pressure effect.
• ### Average Case $(s, t)$-weak tractability of non-homogenous tensor product problems(1802.01829)

Feb. 6, 2018 math.NA
We study $d$-variate problem in the average case setting with respect to a zero-mean Gaussian measure. The covariance kernel of this Gaussian measure is a product of univariate kernels and satisfies some special properties. We study $(s, t)$-weak tractability of this multivariate problem, and obtain a necessary and sufficient condition for $s>0$ and $t\in(0,1)$. Our result can apply to the problems with covariance kernels corresponding to Euler and Wiener integrated processes, Korobov kernels, and analytic Korobov kernels.
• ### Average Case tractability of multivariate approximation with Gaussian kernels(1802.01302)

Feb. 5, 2018 math.NA
We study the problem of approximating functions of $d$ variables in the average case setting for the $L_2$ space $L_{2,d}$ with the standard Gaussian weight equipped with a zero-mean Gaussian measure. The covariance kernel of this Gaussian measure takes the form of a Gaussian kernel with non-increasing positive shape parameters $\gamma_j^2$ for $j = 1, 2, \dots, d$. The error of approximation is defined in the norm of $L_{2,d}$. We study the average case error of algorithms that use at most $n$ arbitrary continuous linear functionals. The information complexity $n(\varepsilon, d)$ is defined as the minimal number of linear functionals which are needed to find an algorithm whose average case error is at most $\varepsilon$. We study different notions of tractability or exponentially-convergent tractability (EC-tractability) which the information complexity $n(\varepsilon, d)$ describe how behaves as a function of $d$ and $\varepsilon^{-1}$ or as one of $d$ and $(1+\ln\varepsilon^{-1})$. We find necessary and sufficient conditions on various notions of tractability and EC-tractability in terms of shape parameters. In particular, for any positive $s>0$ and $t\in(0,1)$ we obtain that the sufficient and necessary condition on $\gamma^2_ j$ for which $$\lim_{d+\varepsilon^{-1}\to\infty}\frac{n(\varepsilon,d)}{\varepsilon^{-s}+d^t}=0$$ holds is $$\lim_{j\to \infty}j^{1-t}\gamma_j^2\,\ln^+ \gamma_j^{-2}=0,$$where $\ln^+ x=\max(1,\ln x)$.
• ### Generating Video Descriptions with Topic Guidance(1708.09666)

Sept. 4, 2017 cs.CV, cs.CL
Generating video descriptions in natural language (a.k.a. video captioning) is a more challenging task than image captioning as the videos are intrinsically more complicated than images in two aspects. First, videos cover a broader range of topics, such as news, music, sports and so on. Second, multiple topics could coexist in the same video. In this paper, we propose a novel caption model, topic-guided model (TGM), to generate topic-oriented descriptions for videos in the wild via exploiting topic information. In addition to predefined topics, i.e., category tags crawled from the web, we also mine topics in a data-driven way based on training captions by an unsupervised topic mining model. We show that data-driven topics reflect a better topic schema than the predefined topics. As for testing video topic prediction, we treat the topic mining model as teacher to train the student, the topic prediction model, by utilizing the full multi-modalities in the video especially the speech modality. We propose a series of caption models to exploit topic guidance, including implicitly using the topics as input features to generate words related to the topic and explicitly modifying the weights in the decoder with topics to function as an ensemble of topic-aware language decoders. Our comprehensive experimental results on the current largest video caption dataset MSR-VTT prove the effectiveness of our topic-guided model, which significantly surpasses the winning performance in the 2016 MSR video to language challenge.
• ### Video Captioning with Guidance of Multimodal Latent Topics(1708.09667)

Sept. 2, 2017 cs.CV, cs.CL
• ### Approximation numbers of Sobolev and Gevrey type embeddings on the sphere and on the ball -- Preasymptotics, asymptotics, and tractability(1701.03545)

Jan. 13, 2017 math.CA
In this paper, we investigate optimal linear approximations ($n$-approximation numbers ) of the embeddings from the Sobolev spaces $H^r\ (r>0)$ for various equivalent norms and the Gevrey type spaces $G^{\alpha,\beta}\ (\alpha,\beta>0)$ on the sphere $\Bbb S^d$ and on the ball $\Bbb B^d$, where the approximation error is measured in the $L_2$-norm. We obtain preasymptotics, asymptotics, and strong equivalences of the above approximation numbers as a function in $n$ and the dimension $d$. We emphasis that all equivalence constants in the above preasymptotics and asymptotics are independent of the dimension $d$ and $n$. As a consequence we obtain that for the absolute error criterion the approximation problems $I_d: H^{r}\to L_2$ are weakly tractable if and only if $r>1$, not uniformly weakly tractable, and do not suffer from the curse of dimensionality. We also prove that for any $\alpha,\beta>0$, the approximation problems $I_d: G^{\alpha,\beta}\to L_2$ are uniformly weakly tractable, not polynomially tractable, and quasi-polynomially tractable if and only if $\alpha\ge 1$.
• ### Preasymptotics and asymptotics of approximation numbers of anisotropic Sobolev embeddings(1607.01865)

Aug. 10, 2016 math.NA
In this paper, we obtain the preasymptotic and asymptotic behavior and strong equivalences of the approximation numbers of the embeddings from the anisotropic Sobolev spaces $W_2^{\bf R}(\Bbb T^d)$ to $L_2(\Bbb T^d)$. We also get the preasymptotic behavior of the approximation numbers of the embeddings from the limit spaces $W_2^{\infty}(\Bbb T^d)$ of the anisotropic Sobolev spaces $W_2^{\bf R}(\Bbb T^d)$ to $L_2(\Bbb T^d)$. We show that both the above embedding problems are intractable and do not suffer from the curse of dimensionality.
• ### Electronic structures of transition metal dipnictides $XPn_2$ ($X$=Ta, Nb; $Pn$=P, As, Sb)(1602.02344)

April 26, 2016 cond-mat.mtrl-sci
The electronic structures and topological properties of transition metal dipnictides $XPn_2$ ($X$=Ta, Nb; $Pn$=P, As, Sb) have been systematically studied using first-principles calculations. In addition to small bulk Fermi surfaces, the band anticrossing features near the Fermi level can be identified from band structures without spin-orbit coupling, leading to nodal lines in all these compounds. Inclusion of spin-orbit coupling gaps out these nodal lines leaving only a pair of disentangled electron/hole bands crossing the Fermi level. Therefore, the low energy physics can be in general captured by the corresponding two band model with several isolated small Fermi pockets. Detailed analysis of the Fermi surfaces suggests that the arsenides and NbSb$_2$ are nearly compensated semimetals while the phosphorides and TaSb$_2$ are not. Based on the calculated band parities, the electron and hole bands are found to be weakly topological non-trivial giving rise to surface states. As an example, we presented the surface-direction-dependent band structure of the surfaces states in TaSb$_2$.
• ### Optical Observations of the Type Ia Supernova 2011fe in M101 for Nearly 500 Days(1602.02951)

Feb. 9, 2016 astro-ph.SR, astro-ph.HE
We present well-sampled optical observations of the bright Type Ia supernova (SN~Ia) SN 2011fe in M101. Our data, starting from $\sim16$ days before maximum light and extending to $\sim463$ days after maximum, provide an unprecedented time series of spectra and photometry for a normal SN~Ia. Fitting the early-time rising light curve, we find that the luminosity evolution of SN 2011fe follows a $t^n$ law, with the index $n$ being close to 2.0 in the $VRI$ bands but slightly larger in the $U$ and $B$ bands. Combining the published ultraviolet (UV) and near-infrared (NIR) photometry, we derive the contribution of UV/NIR emission relative to the optical. SN 2011fe is found to have stronger UV emission and reaches its UV peak a few days earlier than other SNe~Ia with similar $\Delta m_{15}(B)$, suggestive of less trapping of high-energy photons in the ejecta. Moreover, the $U$-band light curve shows a notably faster decline at late phases ($t\approx 100$--300 days), which also suggests that the ejecta may be relatively transparent to UV photons. These results favor the notion that SN 2011fe might have a progenitor system with relatively lower metallicity. On the other hand, the early-phase spectra exhibit prominent high-velocity features (HVFs) of O~I $\lambda$7773 and the Ca~II~NIR triplet, but only barely detectable in Si~II~6355. This difference can be caused either by an ionization/temperature effect or an abundance enhancement scenario for the formation of HVFs; it suggests that the photospheric temperature of SN 2011fe is intrinsically low, perhaps owing to incomplete burning during the explosion of the white dwarf.
• ### Semiparametric GEE analysis in partially linear single-index models for longitudinal data(1507.08473)

July 30, 2015 math.ST, stat.TH
In this article, we study a partially linear single-index model for longitudinal data under a general framework which includes both the sparse and dense longitudinal data cases. A semiparametric estimation method based on a combination of the local linear smoothing and generalized estimation equations (GEE) is introduced to estimate the two parameter vectors as well as the unknown link function. Under some mild conditions, we derive the asymptotic properties of the proposed parametric and nonparametric estimators in different scenarios, from which we find that the convergence rates and asymptotic variances of the proposed estimators for sparse longitudinal data would be substantially different from those for dense longitudinal data. We also discuss the estimation of the covariance (or weight) matrices involved in the semiparametric GEE method. Furthermore, we provide some numerical studies including Monte Carlo simulation and an empirical application to illustrate our methodology and theory.
• ### Density Functional plus Dynamical Mean-Field Theory of the Spin-Crossover Molecule Fe(phen)$_2$(NCS)$_2$(1503.03547)

We study the spin-crossover molecule Fe(phen)$_2$(NCS)$_2$ using density functional theory (DFT) plus dynamical mean-field theory, which allows access to observables not attainable with traditional quantum chemical or electronic structure methods. The temperature dependent magnetic susceptibility, electron addition and removal spectra, and total energies are calculated and compared to experiment. We demonstrate that the proper quantitative energy difference between the high-spin and low-spin state, as well as reasonably accurate values of the magnetic susceptibility can be obtained when using realistic interaction parameters. Comparisons to DFT and DFT+U calculations demonstrate that dynamical correlations are critical to the energetics of the low-spin state. Additionally, we elucidate the differences between DFT+U and spin density functional theory (SDFT) plus U methodologies, demonstrating that DFT+U can recover SDFT+U results for an appropriately chosen on-site exchange interaction.
• ### Unified Structured Learning for Simultaneous Human Pose Estimation and Garment Attribute Classification(1404.4923)

Sept. 22, 2014 cs.CV
In this paper, we utilize structured learning to simultaneously address two intertwined problems: human pose estimation (HPE) and garment attribute classification (GAC), which are valuable for a variety of computer vision and multimedia applications. Unlike previous works that usually handle the two problems separately, our approach aims to produce a jointly optimal estimation for both HPE and GAC via a unified inference procedure. To this end, we adopt a preprocessing step to detect potential human parts from each image (i.e., a set of "candidates") that allows us to have a manageable input space. In this way, the simultaneous inference of HPE and GAC is converted to a structured learning problem, where the inputs are the collections of candidate ensembles, the outputs are the joint labels of human parts and garment attributes, and the joint feature representation involves various cues such as pose-specific features, garment-specific features, and cross-task features that encode correlations between human parts and garment attributes. Furthermore, we explore the "strong edge" evidence around the potential human parts so as to derive more powerful representations for oriented human parts. Such evidences can be seamlessly integrated into our structured learning model as a kind of energy function, and the learning process could be performed by standard structured Support Vector Machines (SVM) algorithm. However, the joint structure of the two problems is a cyclic graph, which hinders efficient inference. To resolve this issue, we compute instead approximate optima by using an iterative procedure, where in each iteration the variables of one problem are fixed. In this way, satisfactory solutions can be efficiently computed by dynamic programming. Experimental results on two benchmark datasets show the state-of-the-art performance of our approach.
• ### Type IIn Supernova SN 2010jl: Optical Observations for Over 500 Days After Explosion(1208.6078)

Aug. 30, 2012 astro-ph.SR
We present extensive optical observations of a Type IIn supernova (SN) 2010jl for the first 1.5 years after the discovery. The UBVRI light curves demonstrated an interesting two-stage evolution during the nebular phase, which almost flatten out after about 90 days from the optical maximum. SN 2010jl has one of the highest intrinsic H_alpha luminosity ever recorded for a SN IIn, especially at late phase, suggesting a strong interaction of SN ejecta with the dense circumstellar material (CSM) ejected by the progenitor. This is also indicated by the remarkably strong Balmer lines persisting in the optical spectra. One interesting spectral evolution about SN 2010jl is the appearance of asymmetry of the Balmer lines. These lines can be well decomposed into a narrow component and an intermediate-width component. The intermediate-width component showed a steady increase in both strength and blueshift with time until t ~ 400 days after maximum, but it became less blueshifted at t ~ 500 days when the line profile appeared relatively symmetric again. Owing to that a pure reddening effect will lead to a sudden decline of the light curves and a progressive blueshift of the spectral lines, we therefore propose that the asymmetric profiles of H lines seen in SN 2010jl is unlikely due to the extinction by newly formed dust inside the ejecta, contrary to the explanation by some early studies. Based on a simple CSM-interaction model, we speculate that the progenitor of SN 2010jl may suffer a gigantic mass loss (~ 30-50 M_sun) in a few decades before explosion. Considering a slow moving stellar wind (e.g., ~ 28 km/s) inferred for the preexisting, dense CSM shell and the extremely high mass-loss rate (1-2 M_sun per yr), we suggest that the progenitor of SN 2010jl might have experienced a red supergiant stage and explode finally as a post-red supergiant star with an initial mass above 30-40 M_sun.
• ### Estimation in semi-parametric regression with non-stationary regressors(1205.3324)

May 15, 2012 math.ST, stat.TH
In this paper, we consider a partially linear model of the form $Y_t=X_t^{\tau}\theta_0+g(V_t)+\epsilon_t$, $t=1,...,n$, where $\{V_t\}$ is a $\beta$ null recurrent Markov chain, $\{X_t\}$ is a sequence of either strictly stationary or non-stationary regressors and $\{\epsilon_t\}$ is a stationary sequence. We propose to estimate both $\theta_0$ and $g(\cdot)$ by a semi-parametric least-squares (SLS) estimation method. Under certain conditions, we then show that the proposed SLS estimator of $\theta_0$ is still asymptotically normal with the same rate as for the case of stationary time series. In addition, we also establish an asymptotic distribution for the nonparametric estimator of the function $g(\cdot)$. Some numerical examples are provided to show that our theory and estimation method work well in practice.
• ### Electronic states and magnetic structure at the Co3O4 (110) surface: a first principles study(1201.6013)

Jan. 29, 2012 cond-mat.mtrl-sci
Tricobalt tetraoxide (Co3O4) is an important catalyst and Co3O4(110) is a frequently exposed surface in Co3O4 nanomaterials. We employed Density-functional theory with on-site Coulomb repulsion U term to study the atomic structures, energetics, magnetic and electronic properties of the two possible terminations, A and B, of this surface. These calculations predict A as the stable termination in a wide range of oxygen chemical potentials, consistent with recent experimental observations. The Co3+ ions do not have a magnetic moment in the bulk, but become magnetic at the surface, which leads to surface magnetic orderings different from the one in the bulk. Surface electronic states are present in the lower half of the bulk band gap and cause partial metallization of both surface terminations. These states are responsible for the charge compensation mechanism stabilizing both polar terminations. The computed critical thickness for polarity compensation is 4 layers.
• ### Electronic structure and bonding properties of cobalt oxide in the spinel structure(1104.4383)

April 22, 2011 cond-mat.mtrl-sci
The spinel cobalt oxide Co3O4 is a magnetic semiconductor containing cobalt ions in Co2+ and Co3+ oxidation states. We have studied the electronic, magnetic and bonding properties of Co3O4 using density functional theory (DFT) at the Generalized Gradient Approximation (GGA), GGA+U, and PBE0 hybrid functional levels. The GGA correctly predicts Co3O4 to be a semiconductor, but severely underestimates the band gap. The GGA+U band gap (1.96 eV) agrees well with the available experimental value (~ 1.6 eV), whereas the band gap obtained using the PBE0 hybrid functional (3.42 eV) is strongly overestimated. All the employed exchange-correlation functionals predict 3 unpaired d electrons on the Co2+ ions, in agreement with crystal field theory, but the values of the magnetic moments given by GGA+U and PBE0 are in closer agreement with the experiment than the GGA value, indicating a better description of the cobalt localized d states. Bonding properties are studied by means of Maximally Localized Wannier Functions (MLWFs). We find d-type MLWFs on the cobalt ions, as well as Wannier functions with the character of sp3d bonds between cobalt and oxygen ions. Such hybridized bonding states indicate the presence of a small covalent component in the primarily ionic bonding mechanism of this compound.
• ### Quasispecies distribution of Eigen model(cond-mat/0608330)

Aug. 15, 2006 cond-mat.stat-mech
We study sharp peak landscapes (SPL) of Eigen model from a new perspective about how the quasispecies distribute in the sequence space. To analyze the distribution more carefully, we bring forth two tools. One tool is the variance of Hamming distance of the sequences at a given generation. It not only offers us a different avenue for accurately locating the error threshold and illustrates how the configuration of the distribution varies with copying fidelity $q$ in the sequence space, but also divides the copying fidelity into three distinct regimes. The other tool is the similarity network of a certain Hamming distance $d_{0}$, by which we can get a visual and in-depth result about how the sequences distribute. We find that there are several local optima around the center (global optimum) in the distribution of the sequences reproduced near the threshold. Furthermore, it is interesting that the distribution of clustering coefficient $C(k)$ follows lognormal distribution and the curve of clustering coefficient $C$ of the network versus $d_{0}$ appears as linear behavior near the threshold.
• ### Charge Transfer Induced Polarity Switching in Carbon Nanotube Transistors(cond-mat/0603322)

March 13, 2006 cond-mat.mtrl-sci
We probed the charge transfer interaction between the amine-containing molecules: hydrazine, polyaniline and aminobutyl phosphonic acid, and carbon nanotube field effect transistors (CNTFETs). We successfully converted p-type CNTFETs to n-type and drastically improved the device performance in both the ON- and OFF- transistor states utilizing hydrazine as dopant. We effectively switched the transistor polarity between p- and n- type by accessing different oxidation states of polyaniline. We also demonstrated the flexibility of modulating the threshold voltage (Vth) of a CNTFET by engineering various charge-accepting and -donating groups in the same molecule.
• ### Self-aligned carbon nanotube transistors with charge transfer doping(cond-mat/0511039)

Nov. 2, 2005 cond-mat.mtrl-sci
This letter reports a charge transfer p-doping scheme which utilizes one-electron oxidizing molecules to obtain stable, unipolar carbon nanotube transistors with a self-aligned gate structure. This doping scheme allows one to improve carrier injection, tune the threshold voltage Vth, and enhance the device performance in both the ON- and OFF- transistor states. Specifically, the nanotube transistor is converted from ambipolar to unipolar, the device drive current is increased by 2-3 orders of magnitude, the device OFF current is suppressed and an excellent Ion/Ioff ratio of six order of magnitude is obtained. The important role played by metal-nanotube contacts modification through charge transfer is demonstrated.