We investigate the problem of incorporating higher-level symbolic score-like information into Automatic Music Transcription (AMT) systems to improve their performance. We use recurrent neural networks (RNNs) and their variants as music language models (MLMs) and present a generative architecture for combining these models with predictions from a frame level acoustic classifier. We also compare different neural network architectures for acoustic modeling. The proposed model computes a distribution over possible output sequences given the acoustic input signal and we present an algorithm for performing a global search for good candidate transcriptions. The performance of the proposed model is evaluated on piano music from the MAPS dataset and we observe that the proposed model consistently outperforms existing transcription methods.
https://arxiv.org/abs/1411.1623
The depth distribution of the transport properties as well as the temperature dependence of the low field magneto-conductance for several c-axis oriented GaN nanowall network samples grown with different average wall-widths are investigated. Magneto-conductance recorded at low temperatures shows clear signature of weak localization effect in all nanowall samples studied here. The scattering mean free path and the phase coherence time, are extracted from the magneto-conductance profile. Electron mobility estimated from scattering mean free path is found to be comparable with those estimated previously from room temperature conductivity data for these samples [Appl. Phys. Lett. 101, 132109 (2012); AIP Conf. Proc. 1583, 252 (2014)], confirming independently the substantial mobility enhancement in these nanowalls as compared to bulk. Our study furthermore reveals that the high electron mobility region extends down to several hundreds of nanometer below the tip of the walls. Like mobility, phase coherence length is found to increase with the reduction of the average wall width. Interestingly, for samples with lower values of the average wall width, phase coherence length is estimated to be as high as 60 micron, which is much larger than those reported for GaN/AlGaN heterostructure based two dimensional electron gas (2DEG) systems.
https://arxiv.org/abs/1411.0366
A major challenge in scaling object detection is the difficulty of obtaining labeled images for large numbers of categories. Recently, deep convolutional neural networks (CNNs) have emerged as clear winners on object classification benchmarks, in part due to training with 1.2M+ labeled classification images. Unfortunately, only a small fraction of those labels are available for the detection task. It is much cheaper and easier to collect large quantities of image-level labels from search engines than it is to collect detection data and label it with precise bounding boxes. In this paper, we propose Large Scale Detection through Adaptation (LSDA), an algorithm which learns the difference between the two tasks and transfers this knowledge to classifiers for categories without bounding box annotated data, turning them into detectors. Our method has the potential to enable detection for the tens of thousands of categories that lack bounding box annotations, yet have plenty of classification data. Evaluation on the ImageNet LSVRC-2013 detection challenge demonstrates the efficacy of our approach. This algorithm enables us to produce a >7.6K detector by using available classification data from leaf nodes in the ImageNet tree. We additionally demonstrate how to modify our architecture to produce a fast detector (running at 2fps for the 7.6K detector). Models and software are available at
https://arxiv.org/abs/1407.5035
AlN layers with thicknesses between 2 and 14 nm were grown on Si(111) substrates by molecular beam epitaxy. The effect of the AlN layer thickness on the morphology and nucleation time of spontaneously formed GaN nanowires (NWs) was investigated by scanning electron microscopy and line-of-sight quadrupole mass spectrometry, respectively. We observed that the alignment of the NWs grown on these layers improves with increasing layer thickness while their nucleation time decreases. Our results show that 4 nm is the smallest thickness of the AlN layer that allows the growth of well-aligned NWs with short nucleation time. Such an AlN buffer layer was successfully employed, together with a patterned SiOx mask, for the selective-area growth (SAG) of vertical GaN NWs. In addition, we fabricated light-emitting diodes (LEDs) from NW ensembles that were grown by means of self-organization phenomena on bare and on AlN-buffered Si substrates. A careful characterization of the optoelectronic properties of the two devices showed that the performance of NW-LEDs on bare and AlN-buffered Si is similar. Electrical conduction across the AlN buffer is facilitated by a high number of grain boundaries that were revealed by transmission electron microscopy. These results demonstrate that grainy AlN buffer layers on Si are compatible both with the SAG of GaN NWs and LED operation. Therefore, this study is a first step towards the fabrication of LEDs on Si substrates based on homogeneous NW ensembles.
https://arxiv.org/abs/1410.7546
High-resolution cameras produce huge volume of high quality images everyday. It is extremely challenging to store, share and especially search those huge images, for which increasing number of cloud services are presented to support such functionalities. However, images tend to contain rich sensitive information (\eg, people, location and event), and people’s privacy concerns hinder their readily participation into the services provided by untrusted third parties. In this work, we introduce PIC: a Privacy-preserving large-scale Image search system on Cloud. Our system enables efficient yet secure content-based image search with fine-grained access control, and it also provides privacy-preserving image storage and sharing among users. Users can specify who can/cannot search on their images when using the system, and they can search on others’ images if they satisfy the condition specified by the image owners. Majority of the computationally intensive jobs are outsourced to the cloud side, and users only need to submit the query and receive the result throughout the entire image search. Specially, to deal with massive images, we design our system suitable for distributed and parallel computation and introduce several optimizations to further expedite the search process. We implement a prototype of PIC including both cloud side and client side. The cloud side is a cluster of computers with distributed file system (Hadoop HDFS) and MapReduce architecture (Hadoop MapReduce). The client side is built for both Windows OS laptops and Android phones. We evaluate the prototype system with large sets of real-life photos. Our security analysis and evaluation results show that PIC successfully protect the image privacy at a low cost of computation and communication.
https://arxiv.org/abs/1410.6593
With the proliferation of mobile devices, cloud-based photo sharing and searching services are becoming common due to the mobile devices’ resource constrains. Meanwhile, there is also increasing concern about privacy in photos. In this work, we present a framework \ourprotocolNSP, which enables cloud servers to provide privacy-preserving photo sharing and search as a service to mobile device users. Privacy-seeking users can share their photos via our framework to allow only their authorized friends to browse and search their photos using resource-bounded mobile devices. This is achieved by our carefully designed architecture and novel outsourced privacy-preserving computation protocols, through which no information about the outsourced photos or even the search contents (including the results) would be revealed to the cloud servers. Our framework is compatible with most of the existing image search technologies, and it requires few changes to the existing cloud systems. The evaluation of our prototype system with 31,772 real-life images shows the communication and computation efficiency of our system.
https://arxiv.org/abs/1410.6589
Object detection performance, as measured on the canonical PASCAL VOC dataset, has plateaued in the last few years. The best-performing methods are complex ensemble systems that typically combine multiple low-level image features with high-level context. In this paper, we propose a simple and scalable detection algorithm that improves mean average precision (mAP) by more than 30% relative to the previous best result on VOC 2012—achieving a mAP of 53.3%. Our approach combines two key insights: (1) one can apply high-capacity convolutional neural networks (CNNs) to bottom-up region proposals in order to localize and segment objects and (2) when labeled training data is scarce, supervised pre-training for an auxiliary task, followed by domain-specific fine-tuning, yields a significant performance boost. Since we combine region proposals with CNNs, we call our method R-CNN: Regions with CNN features. We also compare R-CNN to OverFeat, a recently proposed sliding-window detector based on a similar CNN architecture. We find that R-CNN outperforms OverFeat by a large margin on the 200-class ILSVRC2013 detection dataset. Source code for the complete system is available at this http URL.
https://arxiv.org/abs/1311.2524
Salient object detection has been attracting a lot of interest, and recently various heuristic computational models have been designed. In this paper, we formulate saliency map computation as a regression problem. Our method, which is based on multi-level image segmentation, utilizes the supervised learning approach to map the regional feature vector to a saliency score. Saliency scores across multiple levels are finally fused to produce the saliency map. The contributions lie in two-fold. One is that we propose a discriminate regional feature integration approach for salient object detection. Compared with existing heuristic models, our proposed method is able to automatically integrate high-dimensional regional saliency features and choose discriminative ones. The other is that by investigating standard generic region properties as well as two widely studied concepts for salient object detection, i.e., regional contrast and backgroundness, our approach significantly outperforms state-of-the-art methods on six benchmark datasets. Meanwhile, we demonstrate that our method runs as fast as most existing algorithms.
https://arxiv.org/abs/1410.5926
We report a systematic study of p-type polarization induced doping in graded AlGaN nanowire light emitting diodes grown on silicon wafers by plasma-assisted molecular beam epitaxy. The composition gradient in the p-type base is varied in a set of samples from 0.7 %Al/nm to 4.95 %Al/nm corresponding to negative bound polarization charge densities of 2.2x10^18 cm^-3 to 1.6x10^19 cm^-3. Capacitance measurements and energy band modeling reveal that for gradients greater than or equal to 1.30 %Al/nm, the deep donor concentration is negligible and free hole concentrations roughly equal to the bound polarization charge density are achieved up to 1.6x10^19 cm^-3 at a gradient of 4.95 %Al/nm. Accurate grading lengths in the p- and n-side of the pn-junction are extracted from scanning transmission electron microscopy images and are used to support energy band calculation and capacitance modeling. These results demonstrate the robust nature of p-type polarization doping in nanowires and put an upper bound on the magnitude of deep donor compensation.
https://arxiv.org/abs/1410.5765
JavaScript engines inside modern browsers are capable of running sophisticated multi-player games, rendering impressive 3D scenes, and supporting complex, interactive visualizations. Can this processing power be harnessed for information retrieval? This paper explores the feasibility of building a JavaScript search engine that runs completely self-contained on the client side within the browser—this includes building the inverted index, gathering terms statistics for scoring, and performing query evaluation. The design takes advantage of the IndexDB API, which is implemented by the LevelDB key-value store inside Google’s Chrome browser. Experiments show that although the performance of the JavaScript prototype falls far short of the open-source Lucene search engine, it is sufficiently responsive for interactive applications. This feasibility demonstration opens the door to interesting applications in offline and private search across multiple platforms as well as hybrid split-execution architectures whereby clients and servers collaboratively perform query evaluation. One possible future scenario is the rise of an online search marketplace in which commercial search engine companies and individual users participate as rational economic actors, balancing privacy, resource usage, latency, and other factors based on customizable utility profiles.
https://arxiv.org/abs/1410.4500
Recent studies suggest that piezoelectric polarization can play an important role in determining the electronic and optical properties of nanoscale nitride heterostructures. Among a few models available, recent first-principles calculations performed by Prodhomme et al. provide a simple yet accurate description of linear and nonlinear piezoelectric coefficients in reduced dimensionality structures having wurtzite crystal symmetry. In this paper, first, within a fully atomistic VFF-sp3s* tight-binding framework, we employ the model proposed by Prodhomme et al. to evaluate the importance of nonlinear piezoelectricity on the single-particle electronic states and interband optical transitions in a recently reported hexagon shaped In0.25Ga0.75N/GaN disk-in-wire LED. The microscopically determined transition parameters are then incorporated into a TCAD toolkit to investigate how atomicity and the net polarization field affect the internal quantum efficiency of the LED and lead to a degraded efficiency droop characteristic
https://arxiv.org/abs/1303.6650
We investigate the effect of the p-type top contact on the optoelectronic characteristics of light emitting diodes (LEDs) based on (In,Ga)N/GaN nanowire (NW) ensembles grown by molecular beam epitaxy on Si substrates. We compare devices fabricated with either Ni/Au or indium tin oxide (ITO) top contact. The NW-LEDs with ITO exhibit a number density of NWs emitting electroluminescence about ten times higher, significantly lower turn-on voltage and series resistance, and a relative external quantum efficiency more than one order of magnitude higher than the sample with Ni/Au. These results show that limitations in the performance of such devices reported so far can be overcome by improving the p-type top-contact.
https://arxiv.org/abs/1410.3709
We present a unified framework to describe lattice gauge theories by means of tensor networks: this framework is efficient as it exploits the high amount of local symmetry content native of these systems describing only the gauge invariant subspace. Compared to a standard tensor network description, the gauge invariant one allows to speed-up real and imaginary time evolution of a factor that is up to the square of the dimension of the link variable. The gauge invariant tensor network description is based on the quantum link formulation, a compact and intuitive formulation for gauge theories on the lattice, and it is alternative to and can be combined with the global symmetric tensor network description. We present some paradigmatic examples that show how this architecture might be used to describe the physics of condensed matter and high-energy physics systems. Finally, we present a cellular automata analysis which estimates the gauge invariant Hilbert space dimension as a function of the number of lattice sites and that might guide the search for effective simplified models of complex theories.
https://arxiv.org/abs/1404.7439
Untangling of structures like ropes and wires by autonomous robots can be useful in areas such as personal robotics, industries and electrical wiring & repairing by robots. This problem can be tackled by using computer vision system in robot. This paper proposes a computer vision based method for analyzing visual data acquired from camera for perceiving the overlap of wires, ropes, hoses i.e. detecting tangles. Information obtained after processing image according to the proposed method comprises of position of tangles in tangled object and which wire passes over which wire. This information can then be used to guide robot to untangle wire/s. Given an image, preprocessing is done to remove noise. Then edges of wire are detected. After that, the image is divided into smaller blocks and each block is checked for wire overlap/s and finding other relevant information. TANGLED-100 dataset was introduced, which consists of images of tangled linear deformable objects. Method discussed in here was tested on the TANGLED-100 dataset. Accuracy achieved during experiments was found to be 74.9%. Robotic simulations were carried out to demonstrate the use of the proposed method in applications of robot. Proposed method is a general method that can be used by robots working in different situations.
https://arxiv.org/abs/1405.4802
We propose a scheme of an optical amplifier based on GaN and ZnO waveguides operating in the regime of strong coupling between photonic modes and excitonic resonances. Amplification of the guided exciton-polaritons is obtained by stimulated scattering from the excitonic reservoir, which is found to be fast enough compared with the large velocity of the guided polariton modes. We analyze the device parameters at different temperatures. We find that an 80 $\mu$m-long amplifier can provide a gain of 10dB at room temperature, being supplied by 5 mA current in the $cw$ regime.
https://arxiv.org/abs/1410.2735
Here, we report an alternative route to achieve two dimensional electron gas (2DEG) in a semiconductor structure. It has been shown that charge accumulation on the side facets can lead to the formation of 2DEG in a network of c-axis oriented wedge-shaped GaN nanowalls grown on c-plane sapphire substrate. Our study reveals that negative charges on the side-facets pushes the electron cloud inward resulting in the formation of 2DEG in the central plane parallel to the wall height. This confinement is evidenced from several orders of magnitude enhancement of electron mobility as compared to bulk, observation of weak localization effect in low temperature magneto-transport studies as well as the reduction of both the elastic and inelastic scattering rates with the average width of the walls. Importantly, the phase coherence length has been found to be as high as 20 {\mu}m, which makes the system potentially interesting for spin-tronics. Schrodinger and the Poisson equations are solved self-consistently taking into account the surface charge accumulation effect. The result indeed shows the 2D quantum confinement of electrons even for 40 nm of wall-width.
https://arxiv.org/abs/1410.1295
Real-time detection of moving objects involves memorisation of features in the template image and their comparison with those in the test image. At high sampling rates, such techniques face the problems of high algorithmic complexity and component delays. We present a new resistive switching based threshold logic cell which encodes the pixels of a template image. The cell comprises a voltage divider circuit that programs the resistances of the memristors arranged in a single node threshold logic network and the output is encoded as a binary value using a CMOS inverter gate. When a test image is applied to the template-programmed cell, a mismatch in the respective pixels is seen as a change in the output voltage of the cell. The proposed cell when compared with CMOS equivalent implementation shows improved performance in area, leakage power, power dissipation and delay.
https://arxiv.org/abs/1410.1267
The Imagenet Large Scale Visual Recognition Challenge (ILSVRC) is the one of the most important big data challenges to date. We participated in the object detection track of ILSVRC 2014 and received the fourth place among the 38 teams. We introduce in our object detection system a number of novel techniques in localization and recognition. For localization, initial candidate proposals are generated using selective search, and a novel bounding boxes regression method is used for better object localization. For recognition, to represent a candidate proposal, we adopt three features, namely, RCNN feature, IFV feature, and DPM feature. Given these features, category-specific combination functions are learned to improve the object recognition rate. In addition, object context in the form of background priors and object interaction priors are learned and applied in our system. Our ILSVRC 2014 results are reported alongside with the results of other participating teams.
https://arxiv.org/abs/1409.6155
We report calculated, electronic and related properties of wurtzite and zinc blende gallium nitrides (w-GaN, zb-GaN). We employed a local density approximation (LDA) potential and the linear combination of atomic orbital (LCAO) formalism. The implementation of this formalism followed the Bagayoko, Zhao, and Williams (BZW) method, as enhanced by Ekuma and Franklin (BZW-EF). The calculated electronic and related properties, for both structures of GaN, are in good agreement with corresponding, experimental data, unlike results from most previous ab initio calculations utilizing a density functional theory (DFT) potential. These results include the electronic energy bands, the total and partial densities of states (DOS and pDOS), and effective masses for both structures. The calculated band gap of 3.29 eV, for w-GaN, is in agreement with experiment and is an average of 1.0 eV larger than most previous ab-initio DFT results. Similarly, the calculated band gap of zb-GaN of 2.9 eV, for a room temperature lattice constant, is the ab-initio DFT result closest to the experimental value.
https://arxiv.org/abs/1410.0984
During the earliest phase of architectural design process, practitioners after analyzing the client’s design program, legal requirements, topographic constraints, and preferences synthesize these requirements into architectural floor plan drawings. Design decisions taken in this phase may significantly contribute to the building performance. On account of this reason, it is important to estimate and compare alternative solutions, when it is still manageable to change the building design. The authors have been developing a prototype tool to assist architects during this initial design phase. It is made up of two algorithms. The first algorithm generates alternative floor plans according to the architect’s preferences and requirements, and the client’s design program. It consists in one evolutionary strategy approach enhanced with local search technique to allocate rooms on several levels in the two-dimensional space. The second algorithm evaluates, ranks, and optimizes those floor plans according to thermal performance criteria. The prototype tool is coupled with dynamic simulation program, which estimates the thermal behavior of each solution. A sequential variable optimization is used to change several geometric values of different architectural elements in the floor plans to explore the improvement potential. In the present communication, the two algorithms are used in an iterative process to generate and optimize the thermal performance of alternative floor plans. In the building simulation specifications of EnergyPlus program, the airflow network model has been used in order to adequately model the air infiltration and the airflows through indoor spaces. A case study of a single-family house with three rooms in a single level is presented.
https://arxiv.org/abs/1410.0948
Atomic ensembles are effective memory nodes for quantum communication network due to the long coherence time and the collective enhancement effect for the nonlinear interaction between an ensemble and a photon. Here we investigate the possibility of achieving the entanglement distillation for nonlocal atomic ensembles by the input-output process of a single photon as a result of cavity quantum electrodynamics. We give an optimal entanglement concentration protocol (ECP) for two-atomic-ensemble systems in a partially entangled pure state with known parameters and an efficient ECP for the systems in an unknown partially entangled pure state with a nondestructive parity-check detector (PCD). For the systems in a mixed entangled state, we introduce an entanglement purification protocol with PCDs. These entanglement distillation protocols have high fidelity and efficiency with current experimental techniques, and they are useful for quantum communication network with atomic-ensemble memories.
https://arxiv.org/abs/1403.4660
Metric learning has attracted a lot of interest over the last decade, but the generalization ability of such methods has not been thoroughly studied. In this paper, we introduce an adaptation of the notion of algorithmic robustness (previously introduced by Xu and Mannor) that can be used to derive generalization bounds for metric learning. We further show that a weak notion of robustness is in fact a necessary and sufficient condition for a metric learning algorithm to generalize. To illustrate the applicability of the proposed framework, we derive generalization results for a large family of existing metric learning algorithms, including some sparse formulations that are not covered by previous results.
http://arxiv.org/abs/1209.1086
Empirical temporal networks display strong heterogeneities in their dynamics, which profoundly affect processes taking place on these networks, such as rumor and epidemic spreading. Despite the recent wealth of data on temporal networks, little work has been devoted to the understanding of how such heterogeneities can emerge from microscopic mechanisms at the level of nodes and links. Here we show that long-term memory effects are present in the creation and disappearance of links in empirical networks. We thus consider a simple generative modeling framework for temporal networks able to incorporate these memory mechanisms. This allows us to study separately the role of each of these mechanisms in the emergence of heterogeneous network dynamics. In particular, we show analytically and numerically how heterogeneous distributions of contact durations, of inter-contact durations and of numbers of contacts per link emerge. We also study the individual effect of heterogeneities on dynamical processes, such as the paradigmatic Susceptible-Infected epidemic spreading model. Our results confirm in particular the crucial role of the distributions of inter-contact durations and of the numbers of contacts per link.
https://arxiv.org/abs/1409.1805
Context. Detecting and characterizing circumstellar dust is a way to study the architecture and evolution of planetary systems. Cold dust in debris disks only traces the outer regions. Warm and hot exozodiacal dust needs to be studied in order to trace regions close to the habitable zone. Aims. We aim to determine the prevalence and to constrain the properties of hot exozodiacal dust around nearby main-sequence stars. Methods. We search a magnitude limited (H < 5) sample of 92 stars for bright exozodiacal dust using our VLTI visitor instrument PIONIER in the H-band. We derive statistics of the detection rate with respect to parameters such as the stellar spectral type and age or the presence of a debris disk in the outer regions of the systems. We derive more robust statistics by combining our sample with the results from our CHARA/FLUOR survey in the K-band. In addition, our spectrally dispersed data allows us to put constraints on the emission mechanism and the dust properties in the detected systems. Results. We find an over-all detection rate of bright exozodiacal dust in the H-band of 11% (9 out of 85 targets) and three tentative detections. The detection rate decreases from early type to late type stars and increases with the age of the host star. We do not confirm the tentative correlation between the presence of cold and hot dust found in our earlier analysis of the FLUOR sample alone. Our spectrally dispersed data suggest that either the dust is extremely hot or the emission is dominated by the scattered light in most cases. The implications of our results for the target selection of future terrestrial planet finding missions using direct imaging are discussed.
https://arxiv.org/abs/1409.6143
Multiwavelength variability of blazars offers indirect insight into their powerful engines and on the mechanisms through which energy is propagated from the centre down the jet. The BL Lac object Mkn 421 is a TeV emitter, a bright blazar at all wavelengths, and therefore an excellent target for variability studies. Mkn 421 was observed by INTEGRAL and Fermi-LAT in an active state on 16-21 April 2013. Well sampled optical, soft, and hard X-ray light curves show the presence of two flares. The average flux in the 20-100 keV range is 9.1e-11 erg/s/cm2 (~4.5 mCrab) and the nuclear average apparent magnitude, corrected for Galactic extinction, is V ~12.2. In the time-resolved X-ray spectra (3.5-60 keV), which are described by broken power laws and, marginally better, by log-parabolic laws, we see a hardening that correlates with flux increase, as expected in refreshed energy injections in a population of electrons that later cool via synchrotron radiation. The hardness ratios between the JEM-X fluxes in two different bands and between the JEM-X and IBIS/ISGRI fluxes confirm this trend. During the observation, the variability level increases monotonically from the optical to the hard X-rays, while the large LAT errors do not allow a significant assessment of the MeV-GeV variability. The cross-correlation analysis during the onset of the most prominent flare suggests a monotonically increasing delay of the lower frequency emission with respect to that at higher frequency, with a maximum time-lag of about 70 minutes, that is however not well constrained. The spectral energy distributions from the optical to the TeV domain are satisfactorily described by homogeneous models of blazar emission based on synchrotron radiation and synchrotron self-Compton scattering, except in the state corresponding to the LAT softest spectrum and highest flux.
https://arxiv.org/abs/1307.0558
After the first successful LHC run in 2010-2012, plans are actively advancing for a series of upgrades leading eventually to about above times the design-luminosity in about ten years. The larger luminosity will allow to perform precise measurements of the just discovered Higgs boson and to continue searching for new physics beyond the Standard Model. Coping with the high instantaneous and integrated luminosity will be a great challenge for the ATLAS detector and will require changes in most of the subsystems, specially those at low radii and large pseudorapidity, as well as in its trigger architecture. Plans to consolidate and, whenever possible, to improve the physics performance of the current detector over the next decade are summarized in this paper.
https://arxiv.org/abs/1409.5002
In practical Bayesian optimization, we must often search over structures with differing numbers of parameters. For instance, we may wish to search over neural network architectures with an unknown number of layers. To relate performance data gathered for different architectures, we define a new kernel for conditional parameter spaces that explicitly includes information about which parameters are relevant in a given structure. We show that this kernel improves model quality and Bayesian optimization results over several simpler baseline kernels.
https://arxiv.org/abs/1409.4011
We investigate the structural and optical properties of spontaneously formed GaN nanowires with different degrees of coalescence. This quantity is determined by an analysis of the cross-sectional area and perimeter of the nanowires obtained by plan-view scanning electron microscopy. X-ray diffraction experiments are used to measure the inhomogeneous strain in the nanowire ensembles as well as the orientational distribution of the nanowires. The comparison of the results obtained for GaN nanowire ensembles prepared on bare Si(111) and AlN buffered 6H-SiC(000-1) reveals that the main source of the inhomogeneous strain is the random distortions caused by the coalescence of adjacent nanowires. The magnitude of the strain inhomogeneity induced by nanowire coalescence is found not to be determined solely by the coalescence degree, but also by the mutual misorientation of the coalesced nanowires. The linewidth of the donor-bound exciton transition in photoluminescence spectra does not exhibit a monotonic increase with the coalescence degree. In contrast, the comparison of the root mean square strain with the linewidth of the donor-bound exciton transition reveals a clear correlation: the higher the strain inhomogeneity, the larger the linewidth.
https://arxiv.org/abs/1409.3773
In this paper, we propose multi-stage and deformable deep convolutional neural networks for object detection. This new deep learning object detection diagram has innovations in multiple aspects. In the proposed new deep architecture, a new deformation constrained pooling (def-pooling) layer models the deformation of object parts with geometric constraint and penalty. With the proposed multi-stage training strategy, multiple classifiers are jointly optimized to process samples at different difficulty levels. A new pre-training strategy is proposed to learn feature representations more suitable for the object detection task and with good generalization capability. By changing the net structures, training strategies, adding and removing some key components in the detection pipeline, a set of models with large diversity are obtained, which significantly improves the effectiveness of modeling averaging. The proposed approach ranked #2 in ILSVRC 2014. It improves the mean averaged precision obtained by RCNN, which is the state-of-the-art of object detection, from $31\%$ to $45\%$. Detailed component-wise analysis is also provided through extensive experimental evaluation.
https://arxiv.org/abs/1409.3505
In this paper, we study a general linear networked system that contains a tunable memory subsystem; that is, it is decoupled from an optical field for state transportation during the storage process, while it couples to the field during the writing or reading process. The input is given by a single photon state or a coherent state in a pulsed light field. We then completely and explicitly characterize the condition required on the pulse shape achieving the perfect state transfer from the light field to the memory subsystem. The key idea to obtain this result is the use of zero-dynamics principle, which in our case means that, for perfect state transfer, the output field during the writing process must be a vacuum. A useful interpretation of the result in terms of the transfer function is also given. Moreover, a four-nodes network composed of atomic ensembles is studied as an example, demonstrating how the input field state is transferred to the memory subsystem and how the input pulse shape to be engineered for perfect memory looks like.
https://arxiv.org/abs/1403.1698
ARTOS is all about creating, tuning, and applying object detection models with just a few clicks. In particular, ARTOS facilitates learning of models for visual object detection by eliminating the burden of having to collect and annotate a large set of positive and negative samples manually and in addition it implements a fast learning technique to reduce the time needed for the learning step. A clean and friendly GUI guides the user through the process of model creation, adaptation of learned models to different domains using in-situ images, and object detection on both offline images and images from a video stream. A library written in C++ provides the main functionality of ARTOS with a C-style procedural interface, so that it can be easily integrated with any other project.
https://arxiv.org/abs/1407.2721
The geometrical and performance scaling of silicon CMOS integrated circuit technology over the past 50 years has enabled many affordable new products for business and consumer applications. Recognizing that Flash is approaching its ultimate physical scaling limits within the next 10 years or so, the global electronics research community has begun an intense search for a new paradigm and technology for extending the functional scaling of memory technologies. Several promising nonvolatile memory concepts have emerged, based on different switching and retention mechanisms from those of Flash memory, e.g., STTRAM, RRAM, PCM and more recently, resistive memories based on carbon, which are the topic of this paper. This paper will introduce into the diverse field of carbon materials by recollecting some effects in carbon that can be used to produce a multiple time switchable, non-volatile unipolar resistive memory with potential high scalability down to atomic dimensions. Carbon-based memory is a non-volatile resistive memory, therefore, the same architectures, circuits, select transistor or diodes like in ReRAM or PCRAM can be considered as implementation. The big advantage of carbon memory might be the high temperature retention of 250 C, which makes it attractive for automotive and harsh conditions. This is a white paper for the ITRS meeting on emerging research devices (ERD) in Albuquerque, New Mexico, on August 25-26, 2014.
https://arxiv.org/abs/1408.4600
We describe a method for visual object detection based on an ensemble of optimized decision trees organized in a cascade of rejectors. The trees use pixel intensity comparisons in their internal nodes and this makes them able to process image regions very fast. Experimental analysis is provided through a face detection problem. The obtained results are encouraging and demonstrate that the method has practical value. Additionally, we analyse its sensitivity to noise and show how to perform fast rotation invariant object detection. Complete source code is provided at this https URL.
https://arxiv.org/abs/1305.4537
Results solving the long standing puzzle regarding the phase diagram and the pressure evolution of the melting temperature Tm(P) of gallium nitride (GaN), the most promising semiconducting material for innovative modern electronic applications, are presented. The analysis is based on (i) studies of the decomposition curve in P-T plane up to challenging P equal 9 GPa, (ii) novel method enabling Tm(P) determination despite the earlier decomposition, and (iii) the pressure invariant parameterization of Tm(P) curve, showing the reversal melting for P greater-than 22 GPa. This is linked to a possible fluid-fluid crossover under extreme pressures and temperatures. The importance of results for the development of GaN based technologies is indicated.
https://arxiv.org/abs/1408.3254
Random walks on networks is the standard tool for modelling spreading processes in social and biological systems. This first-order Markov approach is used in conventional community detection, ranking, and spreading analysis although it ignores a potentially important feature of the dynamics: where flow moves to may depend on where it comes from. Here we analyse pathways from different systems, and while we only observe marginal consequences for disease spreading, we show that ignoring the effects of second-order Markov dynamics has important consequences for community detection, ranking, and information spreading. For example, capturing dynamics with a second-order Markov model allows us to reveal actual travel patterns in air traffic and to uncover multidisciplinary journals in scientific communication. These findings were achieved only by using more available data and making no additional assumptions, and therefore suggest that accounting for higher-order memory in network flows can help us better understand how real systems are organized and function.
https://arxiv.org/abs/1305.4807
The article presents the results of the study of design features of vertical and L-shaped ancient Egyptian sundials. With the help of astronomical methods were developed their models, based on which the reconstruction of a sundial was held. Also, the original scheme is a simple way to fairly precise of measurement of time with them has been developed. Large urgency of the task due to the lack of similar models and schemes to date. Model offered by us, which describes the vertical sundial, is a vertical sundial, with a sloping gnomon, which takes into account latitude of area. It is based on the assumption of the existence in ancient Egypt representations about an hour (and a half hour) of equal duration throughout the day, does not depend on the time of year. Offered by us model is characterized by marking hour lines from 6 to 12 hours after each hour. From 12 to 12.5 hours produced displacement in the markup of hour lines on half an hour, then the markup is repeated every hour. As a consequence, the reconstruction of the vertical sundials, we have developed and proposed a model that describes the design features and operation of the L-shaped sundials of two types. They had to work together with the inclined gnomon, like vertical sundials or directly with vertical sundials. In this case, L-shaped sundials can complement vertical sundials, providing an opportunity to read the caption to hour markers and interpret the indications of vertical sundials because vertical sundials inscription missing. The article also describes explanation of the inscriptions from the tomb of Seti I, long intriguing researchers. It is proved that the inscription contains the length of the intervals between adjacent markers L-shaped sundial of the second type, where the first interval corresponds to a half of hour. Keywords: sundial, model, astronomical methods, archaeoastronomy, ancient Egypt.
文章介绍了垂直和L形埃及日design的设计特点的研究结果。在天文学方法的帮助下，开发了他们的模型，在此基础上重建了日was。另外，原来的方案是一个简单的方法，相当精确的测量时间与他们已经开发。由于迄今缺乏类似的模式和计划，这项任务的紧迫性很大。我们提供的描述垂直日Model的模型是一个垂直的日，，带有一个倾斜的g子，它考虑了面积的纬度。它是基于古埃及代表性存在的假设，一天一个小时（半小时），而不是一年中的时间。我们提供的模型的特点是在每小时之后的6到12小时内标记小时线。在12小时到12.5小时的时间段内，在半个小时的时间里产生了排水量，然后每小时重复一次。因此，对垂直日reconstruction的重建，我们已经开发并提出了一个描述两种L形日design的设计特点和操作的模型。他们必须与倾斜的gnomon一起工作，如垂直的日or或直接与垂直的日ials。在这种情况下，L形的日can可以补充垂直的日，，提供一个机会阅读标题小时标记和解释垂直日indic的迹象，因为垂直的日ins铭文丢失。这篇文章还描述了塞蒂一世墓碑铭的解释，这些长期吸引人的研究人员。证明了题字包含了第二类相邻标记L形日intervals之间间隔的长度，其中第一个间隔对应于半小时。关键词：日su，模型，天文学方法，考古天文学，古埃及
https://arxiv.org/abs/1408.0987
We investigate the origin of the fast recombination dynamics of bound and free excitons in GaN nanowire ensembles by temperature-dependent photoluminescence spectroscopy using both continuous-wave and pulsed excitation. The exciton recombination in the present GaN nanowires is dominated by a nonradiative channel between 10 and 300 K. Furthermore, bound and free excitons in GaN NWs are strongly coupled even at low temperatures resulting in a common lifetime of these states. By solving the rate equations for a coupled two-level system, we show that one cannot, in practice, distinguish whether the nonradiative decay occurs directly via the bound or indirectly via the free state. The nanowire surface and coalescence-induced dislocations appear to be the most obvious candidates for nonradiative defects, and we thus compare the exciton decay times measured for a variety of GaN nanowire ensembles with different surface-to-volume ratio and coalescence degrees. The data are found to exhibit no correlation with either of these parameters, i. e., the dominating nonradiative channel in the GaN nanowires under investigation is neither related to the nanowire surface, nor to coalescence-induced defects for the present samples. Hence, we conclude that nonradiative point defects are the origin of the fast recombination dynamics of excitons in GaN nanowires.
https://arxiv.org/abs/1408.1236
It is clear that the current attempts at using algorithms to create artificial neural networks have had mixed success at best when it comes to creating large networks and/or complex behavior. This should not be unexpected, as creating an artificial brain is essentially a design problem. Human design ingenuity still surpasses computational design for most tasks in most domains, including architecture, game design, and authoring literary fiction. This leads us to ask which the best way is to combine human and machine design capacities when it comes to designing artificial brains. Both of them have their strengths and weaknesses; for example, humans are much too slow to manually specify thousands of neurons, let alone the billions of neurons that go into a human brain, but on the other hand they can rely on a vast repository of common-sense understanding and design heuristics that can help them perform a much better guided search in design space than an algorithm. Therefore, in this paper we argue for a mixed-initiative approach for collaborative online brain building and present first results towards this goal.
https://arxiv.org/abs/1408.0998
The relaxation process of electron spin in systems of electrons interacting with piezoelectric deformation phonons that are mediated through spin-orbit interactions was interpreted from a microscopic point of view using the formula for the electron spin relaxation times derived by a projection-reduction method. The electron spin relaxation times in two polymorphic structures of GaN were calculated. The piezoelectric material constant for the wurtzite structure obtained by a comparison with a previously reported experimental result was P_pe=1.5 x 10^29 eV/m. The temperature and magnetic field dependence of the relaxation times for both wurtzite and zinc-blende structures were similar, but the relaxation times in zinc-blende GaN were smaller and decreased more rapidly with increasing temperature and magnetic field than that in wurtzite GaN. This study also showed that the electron spin relaxation for wurtzite GaN at low density could be explained by the Elliot-Yafet process but not for zinc-blende GaN in the metallic regime.
https://arxiv.org/abs/1408.0554
Basal-plane stacking faults are an important class of optically active structural defects in wurtzite semiconductors. The local deviation from the 2H stacking of the wurtzite matrix to a 3C zinc-blende stacking induces a bound state in the gap of the host crystal, resulting in the localization of excitons. Due to the two-dimensional nature of these planar defects, stacking faults act as quantum wells, giving rise to radiative transitions of excitons with characteristic energies. Luminescence spectroscopy is thus capable of detecting even a single stacking fault in an otherwise perfect wurtzite crystal. This review draws a comprehensive picture of the luminescence properties related to stacking faults in GaN. The emission energies associated with different types of stacking faults as well as factors that can shift these energies are discussed. In this context, the importance of the quantum-confined Stark effect in these zinc-blende/wurtzite heterostructures, which results from the spontaneous polarization of wurtzite GaN, is underlined. This discussion is extended to zinc-blende segments in a wurtzite matrix. Furthermore, other factors affecting the emission energy and linewidth of stacking fault-related peaks as well as results obtained at room temperature are addressed. The considerations presented in this article should be transferable also to other wurtzite semiconductors.
https://arxiv.org/abs/1405.1261
An ERP is a kind of package which consist front end and backend as DBMS like a collection of DBMSs. You can create DBMS to manage one aspect of your business. For example, a publishing house has a database of books that keeps information about books such as Author Name, Title, Translator Name, etc. But this database app only helps enter books data and search them. It doesn’t help them, for example, sell books. They get or develop another DBMS database that has all the Books data plus prices, discount formulas, names of common clients, etc. Now they connect the Books database to Sales database and maybe also the inventory database. Now its DBMS slowly turning into an ERP. They may add payroll database and connect it to this ERP. They may develop sales staff and commissions database and connect it to this ERP and so on. In the traditional Database management system the different databases are used for the various Campuses of the JSPM Group of Education like Wagholi Campus, Tathwade Campus, Narhe Campus, Hadpsar Campuses, Bhavdhan Campus as well as Corporate office at Katraj of same organization so it is not possible to keep different databases for the same so in this paper proposed the use of Integrated Database for the Entire organization using ERP system. The Proposed ERP system applied on the existing Architecture of the JSPM Group; the marginal difference observed in the Databases need to be accessed to generate the same number of Reports when use the Traditional DBMS which end up with improvement in the Functional efficiency of Organizational Architecture.
https://arxiv.org/abs/1407.8515
Traditional convolutional neural networks (CNN) are stationary and feedforward. They neither change their parameters during evaluation nor use feedback from higher to lower layers. Real brains, however, do. So does our Deep Attention Selective Network (dasNet) architecture. DasNets feedback structure can dynamically alter its convolutional filter sensitivities during classification. It harnesses the power of sequential processing to improve classification performance, by allowing the network to iteratively focus its internal attention on some of its convolutional filters. Feedback is trained through direct policy search in a huge million-dimensional parameter space, through scalable natural evolution strategies (SNES). On the CIFAR-10 and CIFAR-100 datasets, dasNet outperforms the previous state-of-the-art model.
https://arxiv.org/abs/1407.3068
Searching for nearby exoplanets with direct imaging is one of the major scientific drivers for both space and ground-based programs. While the second generation of dedicated high-contrast instruments on 8-m class telescopes is about to greatly expand the sample of directly imaged planets, exploring the planetary parameter space to hitherto-unseen regions ideally down to Terrestrial planets is a major technological challenge for the forthcoming decades. This requires increasing spatial resolution and significantly improving high contrast imaging capabilities at close angular separations. Segmented telescopes offer a practical path toward dramatically enlarging telescope diameter from the ground (ELTs), or achieving optimal diameter in space. However, translating current technological advances in the domain of high-contrast imaging for monolithic apertures to the case of segmented apertures is far from trivial. SPEED (the segmented pupil experiment for exoplanet detection) is a new instrumental facility in development at the Lagrange laboratory for enabling strategies and technologies for high-contrast instrumentation with segmented telescopes. SPEED combines wavefront control including precision segment phasing architectures, wavefront shaping using two sequential high order deformable mirrors for both phase and amplitude control, and advanced coronagraphy struggled to very close angular separations (PIAACMC). SPEED represents significant investments and technology developments towards the ELT area and future spatial missions, and will offer an ideal cocoon to pave the road of technological progress in both phasing and high-contrast domains with complex/irregular apertures. In this paper, we describe the overall design and philosophy of the SPEED bench.
https://arxiv.org/abs/1407.6826
In this paper we study the problem of object detection for RGB-D images using semantically rich image and depth features. We propose a new geocentric embedding for depth images that encodes height above ground and angle with gravity for each pixel in addition to the horizontal disparity. We demonstrate that this geocentric embedding works better than using raw depth images for learning feature representations with convolutional neural networks. Our final object detection system achieves an average precision of 37.3%, which is a 56% relative improvement over existing methods. We then focus on the task of instance segmentation where we label pixels belonging to object instances found by our detector. For this task, we propose a decision forest approach that classifies pixels in the detection window as foreground or background using a family of unary and binary tests that query shape and geocentric pose features. Finally, we use the output from our object detectors in an existing superpixel classification framework for semantic scene segmentation and achieve a 24% relative improvement over current state-of-the-art for the object categories that we study. We believe advances such as those represented in this paper will facilitate the use of perception in fields like robotics.
https://arxiv.org/abs/1407.5736
We have measured the donor-bound electron spin dynamics in cubic GaN by time-resolved Kerr rotation experiments. The ensemble electron spin dephasing time in this quantum dot like system characterized by a Bohr radius of 2.5 nm is of the order of 1.5 ns as a result of the interaction with the fluctuating nuclear spins. It increases drastically when an external magnetic field as small as 10 mT is applied. We extract a dispersion of the nuclear hyperfine field {\delta}Bn $\sim$ 4 mT, in agreement with calculations. We also demonstrate for the first time in GaN based systems the optical pumping of nuclear spin yielding the build-up of a significant nuclear polarization.
https://arxiv.org/abs/1407.5540
We present here a fully first-principles method for predicting the atomic structure of interfaces. Our method is based on the {\it ab initio} random structure searching (AIRSS) approach, applied here to treat two dimensional defects. The method relies on repeatedly generating random structures in the vicinity of the interface and relaxing them within the framework of density functional theory (DFT). The method is simple, requiring only a small set of parameters that can be easily connected to the chemistry of the system of interest, and efficient, ideally adapted to high-throughput first-principles calculations on modern parallel architectures. Being first-principles, our method is transferable, an important requirement for a generic computational method for the determination of the structure of interfaces. Results for two structurally and chemically very different interfaces are presented here, grain boundaries in graphene and grain boundaries in strontium titanate (SrTiO$_3$). We successfully find a previously unknown low energy grain boundary structure for the graphene system, as well as recover the previously known higher energy structures. For the SrTiO$_3$ system we study both stoichiometric and non-stoichiometric compositions near the grain boundary and find previously unknown low energy structures for all stoichiometries. We predict that these low energy structures have long-range distortions to the ground state crystal structure emanating into the bulk from the interface.
https://arxiv.org/abs/1407.2153
We observe unusually narrow donor-bound exciton transitions (0.4 meV) in the photoluminescence spectra of GaN nanowire ensembles grown on Si(111) substrates at very high (> 850 degrees Celsius) temperatures. The spectra of these samples reveal a prominent transition of excitons bound to neutral Si impurities which is not observed for samples grown under standard conditions. Motivated by these experimental results, we investigate theoretically the impact of surface-induced internal electric fields on the binding energy of donors by a combined Monte Carlo and envelope function approach. We obtain the ranges of doping and diameter for which the potential is well described using the Poisson equation, where one assumes a spatially homogeneous distribution of dopants. Our calculations also show that surface donors in nanowires with a diameter smaller than 100 nm are ionized when the surface electric field is larger than about 10 kV/cm, corresponding to a doping level higher than 2 x 10^16 cm^-3. This result explains the experimental observation: since the (D+,X) complex is not stable in GaN, surface-donor-bound excitons do not contribute to the photoluminescence spectra of GaN nanowires above a certain doping level, and the linewidth reflects the actual structural perfection of the nanowire ensemble.
https://arxiv.org/abs/1407.4279
Personalized recommender systems rely on each user’s personal usage data in the system, in order to assist in decision making. However, privacy policies protecting users’ rights prevent these highly personal data from being publicly available to a wider researcher audience. In this work, we propose a memory biased random walk model on multilayer sequence network, as a generator of synthetic sequential data for recommender systems. We demonstrate the applicability of the synthetic data in training recommender system models for cases when privacy policies restrict clickstream publishing.
https://arxiv.org/abs/1201.6134
We propose a new architecture for difficult image processing operations, such as natural edge detection or thin object segmentation. The architecture is based on a simple combination of convolutional neural networks with the nearest neighbor search. We focus our attention on the situations when the desired image transformation is too hard for a neural network to learn explicitly. We show that in such situations, the use of the nearest neighbor search on top of the network output allows to improve the results considerably and to account for the underfitting effect during the neural network training. The approach is validated on three challenging benchmarks, where the performance of the proposed architecture matches or exceeds the state-of-the-art.
https://arxiv.org/abs/1406.6558
A Content Addressable Memory (CAM) is a memory primarily designed for high speed search operation. Parallel search scheme forms the basis of CAM, thus power reduction is the challenge associated with a large amount of parallel active circuits. We are presenting a novel algorithm and architecture described as Selective Match-Line Energizer Content Addressable Memory (SMLE-CAM) which energizes only those MLs (Match-Line) whose first three bits are conditionally matched with corresponding first three search bit using special architecture which comprises of novel XNOR-CAM cell and novel XOR-CAM cell. The rest of the CAM chain is followed by NOR-CAM cell. The 256 X 144 bit SMLE-CAM is implemented in TSMC 90 nm technology and its robustness across PVT variation is verified. The post-layout simulation result shows, it has energy metric of 0.115 fJ/bit/search with search time 361.6 ps, the best reported so far. The maximum operating frequency is 1GHz.
https://arxiv.org/abs/1406.7662