Search

Deep neural networks have been one of the dominant machine learning approaches in recent years. Several new network structures are proposed and have better performance than the traditional feedforward neural network structure. Representative ones include the skip connection structure in ResNet and the dense connection structure in DenseNet. However, it still lacks a unified guidance for the neural network structure design. In this paper, we propose the hypothesis that the neural network structure design can be inspired by optimization algorithms and a faster optimization algorithm may lead to a better neural network structure. Specifically, we prove that the propagation in the feedforward neural network with the same linear transformation in different layers is equivalent to minimizing some function using the gradient descent algorithm. Based on this observation, we replace the gradient descent algorithm with the heavy ball algorithm and Nesterov's accelerated gradient descent algorithm, which are faster and inspire us to design new and better network structures. ResNet and DenseNet can be considered as two special cases of our framework. Numerical experiments on CIFAR-10, CIFAR-100 and ImageNet verify the advantage of our optimization algorithm inspired structures over ResNet and DenseNet.

Low temperature growth Pr0.7Ca0.3MnO3 (PCMO) thin film showed high performance in electric field induced resistance switching (RS). To understand the micro-mechanism of RS in Metal/PCMO/Metal devices, structure evolution of PCMO under external electric field monitored inside transmission electron microscope (TEM) were performed. Evolution of the modulation stripe in as-grown PCMO sample was investigated when applying electric field. The new-generated modulation stripe gradually disappeared. These results indicate that oxygen ion migration plays a key role in RS., Comment: This paper has been withdrawn by the author due to some typing errors

Vertically aligned ZnO nanotube arrays fabricated on an ITO substrate are found to exhibit strong persistent photoconductivity (PPC) effect and electrically driven conductance switching behavior, though the latter shows a gradual decay from high conductance state to a low conductance state. Unlike the electrical switching, the PPC cannot be reset or reversed by an electrical pulse. Excitation wavelength dependent conductance measurement indicates the presence of the defect localized states (DLS) ~ 240meV above the valence band edge, in support of the hypothesis that the doubly ionization of these DLS are responsible for the PPC effect., Comment: This paper has been withdrawn

Photoconductivity of single-crystalline selenium nanotubes (SCSNT) under a range of illumination intensities of a 633nm laser is carried out with a novel two terminal device arrangement at room temperature. It's found that SCSNT forms Schottky barriers with the W and Au contacts, and the barrier height is a function of the light intensities. In low illumination regime below 1.46x10E-4 muWmum-2, the Au-Se-W hybrid structure exhibits sharp switch on/off behavior, and the turn-on voltages decrease with increasing illuminating intensities. In the high illumination regime above 7x10E-4 muWmum-2, the device exhibits ohmic conductance with a photoconductivity as high as 0.59Ohmcm-1, significantly higher that reported values for carbon and GaN nanotubes. This finding suggests that SCSNT is potentially a good photo-sensor material as well we a very effective solar cell material., Comment: 12pages including 5 figures, submitted to Nanotechnology

We investigated the spin dynamics of two-dimensional electrons in (001) GaAs/AlGaAs heterostructure using the time resolved Kerr rotation technique under a transverse magnetic field. The in-plane spin lifetime is found to be anisotropic below 150k due to the interference of Rashba and Dresselhaus spin-orbit coupling and D'yakonov-Perel' spin relaxation. The ratio of in-plane spin lifetimes is measured directly as a function of temperature and pump power, showing that the electron density in 2DEG channel strongly affects the Rashba spin-orbit coupling., Comment: 3 pages, 2 figures

We observe the inverse spin Hall effect in a two-dimensional electron gas confined in AlGaAs/InGaAs quantum wells. Specifically, we find that an inhomogeneous spin density induced by the optical injection gives rise an electric current transverse to both the spin polarization and its gradient. The spin Hall conductivity can be inferred from such a measurement through the Einstein relation and the Onsager relation, and is found to have the order of magnitude of $0.5(e^{2}/h)$. The observation is made at the room temperature and in samples with macroscopic sizes, suggesting that the inverse spin Hall effect is a robust macroscopic transport phenomenon., Comment: 5 pages, 3 figures, submitted to PRL

We demonstrate for the first time that the quantum size effect (QSE) plays a competing role along side the classical thermodynamic effect in the shape relaxation of a small metallic island. Together, these effects transforms a lead(Pb) island grown on Si(111) substrate from its initially flattop faceted morphology to a peculiar ring-shape island, a process catalysed by the tip electric field of a scanning tunnelling microscope (STM). We shall show for the first time how QSE affects the relaxation process dynamically. In particular, it leads to a novel strip-flow growth and double-step growth on selective strips of a plateau inside the ring, defined by the substrate steps more than 60?0?3 below. It appears that atoms diffusing on the plateau can clearly (sub i)(deg)sense(sub i)+/- the quantized energy states inside the island and have preferentially attached to regions that further reduces the surface energy as a result of the QSE, limiting its own growth and stabilizing the ring shape. The mechanism proposed here offers a sound explanation for ring shape metal and semiconductor islands observed in other systems as well.