Search

The Insulator-Metal-Transition adjoined by a structural transition of VO2 is induced above room temperature (340 K) by heating or self-heating. A steep resistance-jump of up to five orders of magnitude occurs at this transition in high quality, unstrained single crystals. Insulating domains sliding in the sense of the electric current within the metallic background were found so far exclusively in the current induced mixed metal-insulator phase of VO2 single crystals; it is known for a long time that their uniformity and speed as function of current density are very sensitive to crystal quality. The high energetical cost of domain emission is the focus of our present investigations. In this Communication we report on the surprising behavior of a needle-like VO2 single crystal. Several I-V closed loops traced at room temperature concurrently with video recording of the crystal under the microscope, were followed by R(T) measurements during three slow heating-cooling cycles between room temperature and above 340 K, followed in their turn by an additional set of I-V measurements and video recordings. The results show that the slow cycling through the transition using external heat had a healing effect on the reproducibility of R(T) while increasing the activation energy of conduction in the insulating state and in reducing the damping term in the domains' sliding velocity. The intriguing result of this set of measurements was that the energy cost of the domain emission, was higher after healing than prior to it., Comment: 11 pages, 4 figures

Programming has been an important skill for researchers and practitioners in computer science and other related areas. To learn basic programing skills, a long-time systematic training is usually required for beginners. According to a recent market report, the computer software market is expected to continue expanding at an accelerating speed, but the market supply of qualified software developers can hardly meet such a huge demand. In recent years, the surge of text generation research works provides the opportunities to address such a dilemma through automatic program synthesis. In this paper, we propose to make our try to solve the program synthesis problem from a data mining perspective. To address the problem, a novel generative model, namely EgoCoder, will be introduced in this paper. EgoCoder effectively parses program code into abstract syntax trees (ASTs), where the tree nodes will contain the program code/comment content and the tree structure can capture the program logic flows. Based on a new unit model called Hsu, EgoCoder can effectively capture both the hierarchical and sequential patterns in the program ASTs. Extensive experiments will be done to compare EgoCoder with the state-of-the-art text generation methods, and the experimental results have demonstrated the effectiveness of EgoCoder in addressing the program synthesis problem.

Deep neural network learning can be formulated as a non-convex optimization problem. Existing optimization algorithms, e.g., Adam, can learn the models fast, but may get stuck in local optima easily. In this paper, we introduce a novel optimization algorithm, namely GADAM (Genetic-Evolutionary Adam). GADAM learns deep neural network models based on a number of unit models generations by generations: it trains the unit models with Adam, and evolves them to the new generations with genetic algorithm. We will show that GADAM can effectively jump out of the local optima in the learning process to obtain better solutions, and prove that GADAM can also achieve a very fast convergence. Extensive experiments have been done on various benchmark datasets, and the learning results will demonstrate the effectiveness and efficiency of the GADAM algorithm.

In this paper, we aim at introducing a new machine learning model, namely reconciled polynomial machine, which can provide a unified representation of existing shallow and deep machine learning models. Reconciled polynomial machine predicts the output by computing the inner product of the feature kernel function and variable reconciling function. Analysis of several concrete models, including Linear Models, FM, MVM, Perceptron, MLP and Deep Neural Networks, will be provided in this paper, which can all be reduced to the reconciled polynomial machine representations. Detailed analysis of the learning error by these models will also be illustrated in this paper based on their reduced representations from the function approximation perspective.

In this paper, we propose to provide a general ensemble learning framework based on deep learning models. Given a group of unit models, the proposed deep ensemble learning framework will effectively combine their learning results via a multilayered ensemble model. In the case when the unit model mathematical mappings are bounded, sigmoidal and discriminatory, we demonstrate that the deep ensemble learning framework can achieve a universal approximation of any functions from the input space to the output space. Meanwhile, to achieve such a performance, the deep ensemble learning framework also impose a strict constraint on the number of involved unit models. According to the theoretic proof provided in this paper, given the input feature space of dimension d, the required unit model number will be 2d, if the ensemble model involves one single layer. Furthermore, as the ensemble component goes deeper, the number of required unit model is proved to be lowered down exponentially.

In this paper, we introduce an alternative approach, namely GEN (Genetic Evolution Network) Model, to the deep learning models. Instead of building one single deep model, GEN adopts a genetic-evolutionary learning strategy to build a group of unit models generations by generations. Significantly different from the wellknown representation learning models with extremely deep structures, the unit models covered in GEN are of a much shallower architecture. In the training process, from each generation, a subset of unit models will be selected based on their performance to evolve and generate the child models in the next generation. GEN has significant advantages compared with existing deep representation learning models in terms of both learning effectiveness, efficiency and interpretability of the learning process and learned results. Extensive experiments have been done on diverse benchmark datasets, and the experimental results have demonstrated the outstanding performance of GEN compared with the state-of-the-art baseline methods in both effectiveness of efficiency., Comment: arXiv admin note: substantial text overlap with arXiv:1803.08631

Deep learning, a rebranding of deep neural network research works, has achieved a remarkable success in recent years. With multiple hidden layers, deep learning models aim at computing the hierarchical feature representations of the observational data. Meanwhile, due to its severe disadvantages in data consumption, computational resources, parameter tuning costs and the lack of result explainability, deep learning has also suffered from lots of criticism. In this paper, we will introduce a new representation learning model, namely "Sample-Ensemble Genetic Evolutionary Network" (SEGEN), which can serve as an alternative approach to deep learning models. Instead of building one single deep model, based on a set of sampled sub-instances, SEGEN adopts a genetic-evolutionary learning strategy to build a group of unit models generations by generations. The unit models incorporated in SEGEN can be either traditional machine learning models or the recent deep learning models with a much "narrower" and "shallower" architecture. The learning results of each instance at the final generation will be effectively combined from each unit model via diffusive propagation and ensemble learning strategies. From the computational perspective, SEGEN requires far less data, fewer computational resources and parameter tuning efforts, but has sound theoretic interpretability of the learning process and results. Extensive experiments have been done on several different real-world benchmark datasets, and the experimental results obtained by SEGEN have demonstrated its advantages over the state-of-the-art representation learning models., Comment: 12 pages, 5 figures

Transfer of graphene, grown by Chemical Vapor Deposition (CVD), to a substrate of choice, typically involves deposition of a polymeric layer (typically, poly(methyl methacrylate, PMMA or polydimethylsiloxane, PDMS). These polymers are quite hard to remove without leaving some residues behind. Here we study a transfer of graphene with a protective thin oxide layer. The thin oxide layer is grown by Atomic Deposition Layer (ALD) on the graphene right after the growth stage on Cu foils. One can further aid the oxide-graphene transfer by depositing a very thin polymer layer on top of the composite (much thinner than the usual thickness) following by a more aggressive polymeric removal methods, thus leaving the graphene intact. We report on the nucleation growth process of alumina and hafnia films on the graphene, their resulting strain and on their optical transmission. We suggest that hafnia is a better oxide to coat the graphene than alumina in terms of uniformity and defects., Comment: 13 pgs, 13 figures

Nine of the known vanadium oxides, VO$_{2-1/n}$ (n - a positive or negative integer) with n=2 - 6, 8, 9, $\infty$ and -6, undergo metal-insulator transitions accompanied by structural transitions, at various temperatures T$_{MIT}$ (V$_7$O$_{13}$ is metallic above T=0). Among the persistent efforts to determine the driving force(s) of these transitions, electron-electron (Mott-like) and electron-phonon (Peierls-like) interactions, there were several attempts to find systematics in T$_{MIT}$ as function of n. Here we present an unexpectedly simple and illuminating systematics that holds for positive n: if T$_{MIT}$ is the absolute value of the difference between T$_M$(n) and T$_P$(n), which represent the contributions of electron-electron and electron-phonon interactions, respectively, all data points of T$_M$-T$_P$ versus 1/n lie on, or close to, two simple straight lines; one is T$_M$-T$_P$= T$_{\infty}$(7/n-1) for V$_3$O$_5$, V$_4$O$_7$, V$_5$O$_9$, V$_7$O$_{13}$, V$_8$O$_{15}$, V$_9$O$_{17}$ and VO$_2$ and the other is T$_M$-T$_P$= T$_{\infty}$(3/n-1) for V$_2$O$_3$, V$_6$O$_{11}$ and VO$_2$., Comment: Accepted for publication in Solid State Communication

The layered compound BaCo1-xNixS2-y (0.05

The zero-field conductivities ($\sigma$) of the polycrystaline title materials, are governed by inter-grain transport. In the majority of cases their $\sigma$(T) can be described by the "fluctuation induced tunneling" model. Analysis of the results in terms of this model reveals two remarkable features: 1. For \emph{all} Sr$_2$FeMoO$_6$ samples of various microstructures, the tunneling constant (barrier width $\times$ inverse decay-length of the wave-function) is $\sim$ 2, indicating the existence of an intrinsic insulating boundary layer with a well defined electronic (and magnetic) structure. 2. The tunneling constant for \emph{all} cold-pressed samples decreases linearly with increasing magnetic-moment/formula-unit., Comment: 10 pages, 2 tables, 3 figures

We report on the development of a fast neutron detector using a liquid scintillator doped with enriched Li-6. The lithium was introduced in the form of an aqueous LiCl micro-emulsion with a di-isopropylnaphthalene-based liquid scintillator. A Li-6 concentration of 0.15 % by weight was obtained. A 125 mL glass cell was filled with the scintillator and irradiated with fission-source neutrons. Fast neutrons may produce recoil protons in the scintillator, and those neutrons that thermalize within the detector volume can be captured on the Li-6. The energy of the neutron may be determined by the light output from recoiling protons, and the capture of the delayed thermal neutron reduces background events. In this paper, we discuss the development of this 6Li-loaded liquid scintillator, demonstrate the operation of it in a detector, and compare its efficiency and capture lifetime with Monte Carlo simulations. Data from a boron-loaded plastic scintillator were acquired for comparison. We also present a pulse-shape discrimination method for differentiating between electronic and nuclear recoil events based on the Matusita distance between a normalized observed waveform and nuclear and electronic recoil template waveforms. The details of the measurements are discussed along with specifics of the data analysis and its comparison with the Monte Carlo simulation.

For an epitaxial Pr0.48Ca0.52MnO3 film on NdGaO3, we use transmission electron microscopy to observe a "charge-ordered" superlattice along the in-plane direction a. The same film shows no electrical signatures of charge order. The in-plane electrical anisotropy (rho)a/(rho)c = 28 is constant, and there is no evidence of sliding charge density waves up to the large field of ~10^3 V/cm., Comment: 8 pages, 3 figures

We present new accurate measurements of the differential cross section $\sigma(\theta)$ and the proton analyzing power $A_{y}$ for proton-$^{3}$He elastic scattering at various energies. A supersonic gas jet target has been employed to obtain these low energy cross section measurements. The $\sigma(\theta)$ distributions have been measured at $E_{p}$ = 0.99, 1.59, 2.24, 3.11, and 4.02 MeV. Full angular distributions of $A_{y}$ have been measured at $E_{p}$ = 1.60, 2.25, 3.13, and 4.05 MeV. This set of high-precision data is compared to four-body variational calculations employing realistic nucleon-nucleon (NN) and three-nucleon (3N) interactions. For the unpolarized cross section the agreement between the theoretical calculation and data is good when a $3N$ potential is used. The comparison between the calculated and measured proton analyzing powers reveals discrepancies of approximately 50% at the maximum of each distribution. This is analogous to the existing ``$A_{y}$ Puzzle'' known for the past 20 years in nucleon-deuteron elastic scattering., Comment: 22 pages, 9 figures, to be published in Physical Review C, corrected reference 49

We report on pulsed and DC I-V characteristics of polycrystalline samples of three charge-ordered manganites, Pr_{2/3}Ca_{1/3}MnO_3, Pr_{1/2}Ca_{1/2}MnO_3, Bi_{1/2}Sr_{1/2}MnO_3 and of a double-perovskite Sr_2MnReO_6, in a temperature range where their ohmic resistivity obeys the Efros-Shklovskii variable range hopping relation. For all samples, the DC I(V) exhibits at high currents negative differential resistance and hysteresis, which mask a perfectly ohmic or a moderately nonohmic conductivity obtained by pulsed measurements. This demonstrates that the widely used DC I-V measurements are usually misleading., Comment: 6 pages, 4 figures. Accepted for publication to APL

Recent Wilkinson Microwave Anisotropy Probe (WMAP) measurements have determined the baryon density of the Universe $\Omega_b$ with a precision of about 4%. With $\Omega_b$ tightly constrained, comparisons of Big Bang Nucleosynthesis (BBN) abundance predictions to primordial abundance observations can be made and used to test BBN models and/or to further constrain abundances of isotopes with weak observational limits. To push the limits and improve constraints on BBN models, uncertainties in key nuclear reaction rates must be minimized. To this end, we made new precise measurements of the d(d,p)t and d(d,n)^3He total cross sections at lab energies from 110 keV to 650 keV. A complete fit was performed in energy and angle to both angular distribution and normalization data for both reactions simultaneously. By including parameters for experimental variables in the fit, error correlations between detectors, reactions, and reaction energies were accurately tabulated by computational methods. With uncertainties around 2% +/- 1% scale error, these new measurements significantly improve on the existing data set. At relevant temperatures, using the data of the present work, both reaction rates are found to be about 7% higher than those in the widely used Nuclear Astrophysics Compilation of Reaction Rates (NACRE). These data will thus lead not only to reduced uncertainties, but also to modifications in the BBN abundance predictions., Comment: 15 pages, 11 figures, minor editorial changes

Using the transverse field muon spin relaxation technique we measure the temperature dependence of the magnetic field penetration depth $\lambda$, in the Na$_{x}$CoO$_{2}\cdot y$H$_{2}$O system. We find that $\lambda,$ which is determined by superfluid density $n_{s}$ and the effective mass $m^{\ast}$, is very small and on the edge of the TF-$\mu$SR sensitivity. Nevertheless, the results indicate that the order parameter in this system has nodes and that it obeys the Uemura relation. By comparing $\lambda$ with the normal state electron density we conclude that $m^{\ast}$ of the superconductivity carrier is 70 times larger than the mass of bare electrons.

In this paper we report the results of simulations of a 2D gravity driven, dissipative granular flow through a hopper system. Measurements of impulse distributions P(I) on the simulated system show flow-velocity-invariant behavior of the distribution for impulses larger than the average impulse . For small impulses, however, P(I) decreases significantly with flow velocity, a phenomenon which can be attributed exclusively to collisions between grains undergoing frequent collisions. Visualizations of the system also show that these frequently colliding particles tend to form increasingly large linear clusters as the flow velocity decreases. A model is proposed for the form of P(I), given distributions of cluster size and velocity, which accurately predicts the observed form of the distribution. Thus the impulse distribution provides some insight into the formation and properties of these ``dynamic'' force chains., Comment: 4 pages, 4 figures

Angular distributions of sigma(theta), A_y, iT_11, T_20, T_21, and T_22 have been measured for d-p scattering at E_c.m.=667 keV. This set of high-precision data is compared to variational calculations with the nucleon-nucleon potential alone and also to calculations including a three-nucleon (3N) potential. Agreement with cross-section and tensor analyzing power data is excellent when a 3N potential is used. However, a comparison between the vector analyzing powers reveals differences of approximately 40% in the maxima of the angular distributions which is larger than reported at higher energies for both p-d and n-d scattering. Single-energy phase-shift analyses were performed on this data set and a similar data set at E_c.m.=431.3 keV. The role of the different phase-shift parameters in fitting these data is discussed., Comment: 18 pages, 6 figures

A new measurement of the p-d differential cross section at Ep= 1 MeV has been performed. These new data and older data sets at energies below the deuteron breakup are compared to calculations using the two-nucleon Argonne v18 and the three-nucleon Urbana IX potentials. A quantitative estimate of the capability of these interactions to describe the data is given in terms of a chi^2 analysis. The chi^2 per datum drastically improves when the three-nucleon interaction is included in the Hamiltonian., Comment: 13 pages, 5 figures, to be published in Phys. Rev. C