Apr 28 2015

cs.CV arXiv:1504.07159v1

We propose a new learning-based method for estimating 2D human pose from a single image, using Dual-Source Deep Convolutional Neural Networks (DS-CNN). Recently, many methods have been developed to estimate human pose by using pose priors that are estimated from physiologically inspired graphical models or learned from a holistic perspective. In this paper, we propose to integrate both the local (body) part appearance and the holistic view of each local part for more accurate human pose estimation. Specifically, the proposed DS-CNN takes a set of image patches (category-independent object proposals for training and multi-scale sliding windows for testing) as the input and then learns the appearance of each local part by considering their holistic views in the full body. Using DS-CNN, we achieve both joint detection, which determines whether an image patch contains a body joint, and joint localization, which finds the exact location of the joint in the image patch. Finally, we develop an algorithm to combine these joint detection/localization results from all the image patches for estimating the human pose. The experimental results show the effectiveness of the proposed method by comparing to the state-of-the-art human-pose estimation methods based on pose priors that are estimated from physiologically inspired graphical models or learned from a holistic perspective.

Simulating quantum contextuality with classical systems requires memory. A fundamental yet open question is which is the minimum memory needed and, therefore, the precise sense in which quantum systems outperform classical ones. Here we make rigorous the notion of classically simulating quantum state-independent contextuality (QSIC) in the case of a single quantum system submitted to an infinite sequence of measurements randomly chosen from a finite QSIC set. We obtain the minimum memory classical systems need to simulate arbitrary QSIC sets under the assumption that the simulation should not contain any oracular information. In particular, we show that, while classically simulating two qubits tested with the Peres-Mermin set requires $\log_2 24 \approx 4.585$ bits, simulating a single qutrit tested with the Yu-Oh set requires, at least, $5.740$ bits.

Purification is a powerful technique in quantum physics whereby a mixed quantum state is extended to a pure state on a larger system. This process is not unique, and in systems composed of many degrees of freedom, one natural purification is the one with minimal entanglement. Here we study the entropy of the minimally entangled purification, called the entanglement of purification, in three model systems: an Ising spin chain, conformal field theories holographically dual to Einstein gravity, and random stabilizer tensor networks. We conjecture values for the entanglement of purification in all these models, and we support our conjectures with a variety of numerical and analytical results. We find that such minimally entangled purifications have a number of applications, from enhancing entanglement-based tensor network methods for describing mixed states to elucidating novel aspects of the emergence of geometry from entanglement in the AdS/CFT correspondence.

We extend quantum Stein's lemma in asymmetric quantum hypothesis testing to composite null and alternative hypotheses. As our main result, we show that the asymptotic error exponent for testing convex combinations of quantum states $\rho^{\otimes n}$ against convex combinations of quantum states $\sigma^{\otimes n}$ is given by a regularized quantum relative entropy distance formula. We prove that in general such a regularization is needed but also discuss various settings where our formula as well as extensions thereof become single-letter. This includes a novel operational interpretation of the relative entropy of coherence in terms of hypothesis testing. For our proof, we start from the composite Stein's lemma for classical probability distributions and lift the result to the non-commutative setting by only using elementary properties of quantum entropy. Finally, our findings also imply an improved Markov type lower bound on the quantum conditional mutual information in terms of the regularized quantum relative entropy - featuring an explicit and universal recovery map.

Sep 22 2017

cs.CY arXiv:1709.07387v1

Big data analytics has an extremely significant impact on many areas in all businesses and industries including hospitality. This study aims to guide information technology (IT) professionals in hospitality on their big data expedition. In particular, the purpose of this study is to identify the maturity stage of the big data in hospitality industry in an objective way so that hotels be able to understand their progress, and realize what it will take to get to the next stage of big data maturity through the scores they will receive based on the survey.

Social networks and interactions in social media involve both positive and negative relationships. Signed graphs capture both types of relationships: positive edges correspond to pairs of "friends", and negative edges to pairs of "foes". The edge sign prediction problem, that aims to predict whether an interaction between a pair of nodes will be positive or negative, is an important graph mining task for which many heuristics have recently been proposed [Leskovec 2010]. We model the edge sign prediction problem as follows: we are allowed to query any pair of nodes whether they belong to the same cluster or not, but the answer to the query is corrupted with some probability $0<q<\frac{1}{2}$. Let $\delta=1-2q$ be the bias. We provide an algorithm that recovers all signs correctly with high probability in the presence of noise for any constant gap $\delta$ with $O(\frac{n\log n}{\delta^4})$ queries. Our algorithm uses breadth first search as its main algorithmic primitive. A byproduct of our proposed learning algorithm is the use of $s-t$ paths as an informative feature to predict the sign of the edge $(s,t)$. As a heuristic, we use edge disjoint $s-t$ paths of short length as a feature for predicting edge signs in real-world signed networks. Our findings suggest that the use of paths improves the classification accuracy, especially for pairs of nodes with no common neighbors.

G. Pradels, T. Guinle, G. Thuillier, A. Irbah, J-P. Marcovici, C. Dufour, D. Moreau, C. Noel, M. Dominique, T. Corbard, M. Hadjara, S. Mekaoui, C. Wehrli PICARD is a scientific space mission dedicated to the study of the solar variability origin. A French micro-satellite will carry an imaging telescope for measuring the solar diameter, limb shape and solar oscillations, and two radiometers for measuring the total solar irradiance and the irradiance in five spectral domains, from ultraviolet to infrared. The mission is planed to be launched in 2009 for a 3-year duration. This article presents the PICARD Payload Data Centre, which role is to collect, process and distribute the PICARD data. The Payload Data Centre is a joint project between laboratories, space agency and industries. The Belgian scientific policy office funds the industrial development and future operations under the European Space Agency program. The development is achieved by the SPACEBEL Company. The Belgian operation centre is in charge of operating the PICARD Payload Data Centre. The French space agency leads the development in partnership with the French scientific research centre, which is responsible for providing all the scientific algorithms. The architecture of the PICARD Payload Data Centre (software and hardware) is presented. The software system is based on a Service Oriented Architecture. The host structure is made up of the basic functions such as data management, task scheduling and system supervision including a graphical interface used by the operator to interact with the system. The other functions are mission-specific: data exchange (acquisition, distribution), data processing (scientific and non-scientific processing) and managing the payload (programming, monitoring). The PICARD Payload Data Centre is planned to be operated for 5 years. All the data will be stored into a specific data centre after this period.

Although deep Convolutional Neural Network (CNN) has shown better performance in various machine learning tasks, its application is accompanied by a significant increase in storage and computation. Among CNN simplification techniques, parameter pruning is a promising approach which aims at reducing the number of weights of various layers without intensively reducing the original accuracy. In this paper, we propose a novel progressive parameter pruning method, named Structured Probabilistic Pruning (SPP), which efficiently prunes weights of convolutional layers in a probabilistic manner. Unlike existing deterministic pruning approaches, in which the pruned weights of a well-trained model are permanently eliminated, SPP utilizes the relative importance of weights during training iterations, which makes the pruning procedure more accurate by leveraging the accumulated weight importance. Specifically, we introduce an effective weight competition mechanism to emphasize the important weights and gradually undermine the unimportant ones. Experiments indicate that our proposed method has obtained superior performance on ConvNet and AlexNet compared with existing pruning methods. Our pruned AlexNet achieves 4.0 $\sim$ 8.9x (averagely 5.8x) layer-wise speedup in convolutional layers with only 1.3\% top-5 error increase on the ImageNet-2012 validation dataset. We also prove the effectiveness of our method on transfer learning scenarios using AlexNet.

We study the problem of learning description logic (DL) ontologies in Angluin et al.'s framework of exact learning via queries. We admit membership queries ("is a given subsumption entailed by the target ontology?") and equivalence queries ("is a given ontology equivalent to the target ontology?"). We present three main results: (1) ontologies formulated in (two relevant versions of) the description logic DL-Lite can be learned with polynomially many queries of polynomial size; (2) this is not the case for ontologies formulated in the description logic EL, even when only acyclic ontologies are admitted; and (3) ontologies formulated in a fragment of EL related to the web ontology language OWL 2 RL can be learned in polynomial time. We also show that neither membership nor equivalence queries alone are sufficient in cases (1) and (3).

The arXiv is the most popular preprint repository in the world. Since its inception in 1991, the arXiv has allowed researchers to freely share publication-ready articles prior to formal peer review. The growth and the popularity of the arXiv emerged as a result of new technologies that made document creation and dissemination easy, and cultural practices where collaboration and data sharing were dominant. The arXiv represents a unique place in the history of research communication and the Web itself, however it has arguably changed very little since its creation. Here we look at the strengths and weaknesses of arXiv in an effort to identify what possible improvements can be made based on new technologies not previously available. Based on this, we argue that a modern arXiv might in fact not look at all like the arXiv of today.