_id
stringlengths 40
40
| text
stringlengths 0
10k
|
|---|---|
ced7811f2b694e54e3d96ec5398e4b6afca67fc0 | This paper presents a novel illumination normalization approach for face recognition under varying lighting conditions. In the proposed approach, a discrete cosine transform (DCT) is employed to compensate for illumination variations in the logarithm domain. Since illumination variations mainly lie in the low-frequency band, an appropriate number of DCT coefficients are truncated to minimize variations under different lighting conditions. Experimental results on the Yale B database and CMU PIE database show that the proposed approach improves the performance significantly for the face images with large illumination variations. Moreover, the advantage of our approach is that it does not require any modeling steps and can be easily implemented in a real-time face recognition system. |
30a587efb58f103ff4e42444e191f50a99513f1b | In this paper, we consider learning a Bayesian collaborative filtering model on a shared cluster of commodity machines. Two main challenges arise: (1) How can we parallelize and distribute Bayesian collaborative filtering? (2) How can our distributed inference system handle elasticity events common in a shared, resource managed cluster, including resource ramp-up, preemption, and stragglers? To parallelize Bayesian inference, we adapt ideas from both matrix factorization partitioning schemes used with stochastic gradient descent and stale synchronous programming used with parameter servers. To handle elasticity events we offer a generalization of previous partitioning schemes that gives increased flexibility during system disruptions. We additionally describe two new scheduling algorithms to dynamically route work at runtime. In our experiments, we compare the effectiveness of both scheduling algorithms and demonstrate their robustness to system failure. |
dc821ab3a1a3b49661639da37e980bfd21d3746a | |
558fc9a2bce3d3993a9c1f41b6c7f290cefcf92f | |
4ee9f569e7878ef2b8bac9a563312c7176139828 | We propose a new framework for image recognition by selectively pooling local visual descriptors, and show its superior discriminative power on fine-grained image classification tasks. The representation is based on selecting the most confident local descriptors for nonlinear function learning using a linear approximation in an embedded higher dimensional space. The advantage of our Selective Pooling Vector over the previous state-of-the-art Super Vector and Fisher Vector representations, is that it ensures a more accurate learning function, which proves to be important for classifying details in fine-grained image recognition. Our experimental results corroborate this claim: with a simple linear SVM as the classifier, the selective pooling vector achieves significant performance gains on standard benchmark datasets for various fine-grained tasks such as the CMU Multi-PIE dataset for face recognition, the Caltech-UCSD Bird dataset and the Stanford Dogs dataset for fine-grained object categorization. On all datasets we outperform the state of the arts and boost the recognition rates to 96.4%, 48.9%, 52.0% respectively. |
563e821bb5ea825efb56b77484f5287f08cf3753 | |
6434b7610aa9e7c37fce2258ab459fa99a631748 | In this paper, we propose a novel method for solving single-image super-resolution problems. Given a low-resolution image as input, we recover its high-resolution counterpart using a set of training examples. While this formulation resembles other learning-based methods for super-resolution, our method has been inspired by recent manifold teaming methods, particularly locally linear embedding (LLE). Specifically, small image patches in the lowand high-resolution images form manifolds with similar local geometry in two distinct feature spaces. As in LLE, local geometry is characterized by how a feature vector corresponding to a patch can be reconstructed by its neighbors in the feature space. Besides using the training image pairs to estimate the high-resolution embedding, we also enforce local compatibility and smoothness constraints between patches in the target high-resolution image through overlapping. Experiments show that our method is very flexible and gives good empirical results. |
a21bb8ddfe448e890871e248e196dc0f56000d3a | Image-based models for computer graphics lack resolution independence: they cannot be zoomed much beyond the pixel resolution they were sampled at without a degradation of quality. Interpolating images usually results in a blurring of edges and image details. We describe image interpolation algorithms which use a database of training images to create plausible high-frequency details in zoomed images. Image pre-processing steps allow the use of image detail from regions of the training images which may look quite di erent from the image to be processed. These methods preserve ne details, such as edges, generate believable textures, and can give good results even after zooming multiple octaves. This work may not be copied or reproduced in whole or in part for any commercial purpose. Permission to copy in whole or in part without payment of fee is granted for nonpro t educational and research purposes provided that all such whole or partial copies include the following: a notice that such copying is by permission of Mitsubishi Electric Information Technology Center America; an acknowledgment of the authors and individual contributions to the work; and all applicable portions of the copyright notice. Copying, reproduction, or republishing for any other purpose shall require a license with payment of fee to Mitsubishi Electric Information Technology Center America. All rights reserved. Copyright c Mitsubishi Electric Information Technology Center America, 2001 201 Broadway, Cambridge, Massachusetts 02139 1. First printing, TR2001-30, August, 2001. Egon Pasztor's present address: MIT Media Lab 20 Ames St. Cambridge, MA 02139 Example-based Super-resolution William T. Freeman, Thouis R. Jones, Egon C. Pasztor MitsubishiElectricResearchLaboratories(MERL) 201Broadway Cambridge,MA 02139 (a) (b) Figure 1: An object modelledby traditionalpolygon techniques may lack someof the richnessof real-world objects,but behaves properlyunderzooming,(b). |
6f657378907d2e4ad3167f4c5fb7b52426a7d6cb | |
fd4c6f56c58923294d28dc415e1131a976e8ea93 | Purpose – The purpose of this paper is to explore if six sigma and lean are new methods, or if they are repackaged versions of previously popular methods – total quality management (TQM) and just-in-time (JIT). Design/methodology/approach – The study is based on a critical comparison of lean with JIT and six sigma with TQM, a study of the measure of the publication frequency – the number of academic articles published every year of the previous 30 years – for each topic, and a review of critical success factors (CSF) for change efforts. Findings – The more recent concepts of lean and six sigma have mainly replaced – but not necessarily added to – the concepts of JIT and TQM. lean and six sigma are essentially repackaged versions of the former, and the methods seem to follow the fad (product) life cycle. The literature offers fairly similar and rather general CSF for these methods, e.g. top management support and the importance of communication and information. What seems to be missing, however, is the need for a systemic approach to organizational change and improvement. Practical implications – A prediction is, given the fad or product life cycle phenomenon, that there will be a new method promoted soon, something perhaps already experienced with the borderline preposterous concept of lean six sigma. On the other hand, based on the gap in time between both JIT and lean, and TQM and six sigma – a gap filled by BRP/reengineering – the next method will be process oriented. This paper concludes with the discussion of the need for a process-based approach to organizational improvement efforts. Originality/value – This paper is of value in that it analyzes what lessons can be learnt from organizational change and improvement efforts. The analysis includes a comparison of CSF for any change project before discussing the need for a process (systems) perspective for successful organizational improvement efforts. |
364dc9031ec997a15015099d876d8e49c76a4e9f | In the past several years, social media (e.g., Twitter and Facebook) has been experiencing a spectacular rise and popularity, and becoming a ubiquitous discourse for content sharing and social networking. With the widespread of mobile devices and location-based services, social media typically allows users to share whereabouts of daily activities (e.g., check-ins and taking photos), and thus strengthens the roles of social media as a proxy to understand human behaviors and complex social dynamics in geographic spaces. Unlike conventional spatiotemporal data, this new modality of data is dynamic, massive, and typically represented in stream of unstructured media (e.g., texts and photos), which pose fundamental representation, modeling and computational challenges to conventional spatiotemporal analysis and geographic information science. In this paper, we describe a scalable computational framework to harness massive location-based social media data for efficient and systematic spatiotemporal data analysis. Within this framework, the concept of space-time trajectories (or paths) is applied to represent activity profiles of social media users. A hierarchical spatiotemporal data model, namely a spatiotemporal data cube model, is developed based on collections of space-time trajectories to represent the collective dynamics of social media users across aggregation boundaries at multiple spatiotemporal scales. The framework is implemented based upon a public data stream of Twitter feeds posted on the continent of North America. To demonstrate the advantages and performance of this framework, an interactive flow mapping interface (including both single-source and multiple-source flow mapping) is developed to allow real-time, and interactive visual exploration of movement dynamics in massive location-based social media at multiple scales. ∗Corresponding author Preprint submitted to Computers, Enviroment and Urban Systems September 10, 2014 ar X iv :1 40 9. 28 26 v1 [ cs .S I] 8 S ep 2 01 4 |
b51cdeb74165df0ce32ea3a1a606d28186f03e9d | |
c1e4420ddc71c4962e0ba26287293a25a774fb6e | We consider the task of depth estimation from a single monocular image. We take a supervised learning approach to this problem, in which we begin by collecting a training set of monocular images (of unstructured outdoor environments which include forests, trees, buildings, etc.) and their corresponding ground-truth depthmaps. Then, we apply supervised learning to predict the depthmap as a function of the image. Depth estimation is a challenging problem, since local features alone are insufficient to estimate depth at a point, and one needs to consider the global context of the image. Our model uses a discriminatively-trained Markov Random Field (MRF) that incorporates multiscale localand global-image features, and models both depths at individual points as well as the relation between depths at different points. We show that, even on unstructured scenes, our algorithm is frequently able to recover fairly accurate depthmaps. |
29ad218282cd9d66fe921d49ba066dae32bfa76d | The explosion of interconnected devices and the Internet of Things has triggered new important challenges in the area of internet security, due to the various device vulnerabilities and increased potential for cyber-attacks. This paper touches on the areas of Cybersecurity, intrusion detection, prevention systems and artificial intelligence. Our aim is to create a system capable of understanding, detecting and preventing malicious connections using applied concepts of machine learning. We emphasize the importance of selecting and extracting features that can lead to an accurate decision of classification for malware and intrusion attacks. We propose a solution that combines features that extract correlations from the packet history for the same and different services and hosts, based on the rate of REJ, SYN and ACK flags and connection states, with HTTP features extracted from URI and RESTful methods. Our proposed solution is able to detect network intrusions and botnet communications with a precision of 98.4% on the binary classification problem. |
c8d170a6cf0b773a1db726492793c0426c7bc10a | Real Options Theory is often applied to the valuation of IT investments. The application of Real Options Theory is generally accompanied by a monetary valuation of real options through option pricing models which in turn are based on restrictive assumptions and thus subject to criticism. Therefore, this paper analyzes the application of option pricing models for the valuation of IT investments. A structured literature review reveals the types of IT investments which are valuated with Real Options Theory in scientific literature. These types of IT investments are further investigated and their main characteristics are compared to the restrictive assumptions of traditional option pricing models. This analysis serves as a basis for further discussion on how the identified papers address these assumptions. The results show that a lot of papers do not account for critical assumptions, although it is known that the assumptions are not fulfilled. Moreover, the type of IT investment determines the criticality of the assumptions. Additionally, several extensions or adaptions of traditional option pricing models can be found which provide the possibility of relaxing critical assumptions. Researchers can profit from the results derived in this paper in two ways: First, which assumptions can be critical for various types of IT investments are demonstrated. Second, extensions of option pricing models that relax critical assumptions are introduced. |
675b402bd488aae006ff9a9dd0951797558b2d6d | BACKGROUND
Attention-deficit/hyperactivity disorder (ADHD) is a complex highly comorbid disorder, which can have a huge impact on those with ADHD, their family, and the community around them. ADHD is currently managed using pharmacological and nonpharmacological interventions. However, with advances in technology and an increase in the use of mobile apps, managing ADHD can be augmented using apps specifically designed for this population. However, little is known regarding the suitability and usability of currently available apps.
OBJECTIVE
The aim of this study was to explore the suitability of the top 10 listed apps for children and young people with ADHD and clinicians who work with them. It is hypothesized that mobile apps designed for this population could be more suitably designed for this population.
METHODS
The top 10 listed apps that are specifically targeted toward children and young people with ADHD in the United Kingdom were identified via the Google Play (n=5) and iTunes store (n=5). Interviews were then undertaken with 5 clinicians who specialize in treating this population and 5 children and young people with ADHD themselves, to explore their opinions of the 10 apps identified and what they believe the key components are for apps to be suitable for this population.
RESULTS
Five themes emerged from clinician and young people interviews: the accessibility of the technology, the importance of relating to apps, addressing ADHD symptoms and related difficulties, age appropriateness, and app interaction. Three additional themes emerged from the clinician interviews alone: monitoring symptoms, side effects and app effect on relationships, and the impact of common comorbid conditions. The characteristics of the apps did not appear to match well with the views of our sample.
CONCLUSIONS
These findings suggest that the apps may not be suitable in meeting the complex needs associated with this condition. Further research is required to explore the value of apps for children and young people with ADHD and their families and, in particular, any positive role for apps in the management of ADHD in this age group. A systematic review on how technology can be used to engage this population and how it can be used to help them would be a useful way forward. This could be the platform to begin exploring the use of apps further. |
d16afa6fe546451bb4330eae67b1722823e37e41 | We propose an interactive scheme for segmenting meal images for automated dietary assessment. A smartphone user photographs a meal and marks a few touch points on the resulting image. The segmentation algorithm initializes a set of food segments with the touch points, and grows them using local image features. We evaluate the algorithm with a data set consisting of 300 manually segmented meal images. The precision of segmentation is 0.87, compared with 0.70 for fully automatic segmentation. The results show that the precision of segmentation was significantly improved by incorporating minimal user intervention. |
03a3b5ca18f6482cfee128eb24ddd1a59015fb2d | The bloom filter is a probabilistic data structure that provides a compact representation of a set of elements. To keep false positive probabilities low, the size of the bloom filter must be dimensioned a priori to be linear in the maximum number of keys inserted, with the linearity constant ranging typically from one to few bytes. A bloom filter is most commonly used as an in memory data structure, hence its size is limited by the availability of RAM space on the machine. As datasets have grown over time to Internet scale, so have the RAM space requirements of bloom filters. If sufficient RAM space is not available, we advocate that flash memory may serve as a suitable medium for storing bloom filters, since it is about one-tenth the cost of RAM per GB while still providing access times orders of magnitude faster than hard disk. We present BLOOMFLASH, a bloom filter designed for flash memory based storage, that provides a new dimension of trade off with bloom filter access times to reduce RAM space usage (and hence system cost). The simple design of a single flat bloom filter on flash suffers from many performance bottlenecks, including in-place bit updates that are inefficient on flash and multiple reads and random writes spread out across many flash pages for a single lookup or insert operation. To mitigate these performance bottlenecks, BLOOMFLASH leverages two key design innovations: (i) buffering bit updates in RAM and applying them in bulk to flash that helps to reduce random writes to flash, and (ii) a hierarchical bloom filter design consisting of component bloom filters, stored one per flash page, that helps to localize reads and writes on flash. We use two real-world data traces taken from representative bloom filter applications to drive and evaluate our design. BLOOMFLASH achieves bloom filter access times in the range of few tens of microseconds, thus allowing up to order of tens of thousands operations per sec. |
14d792772da62654f5c568153222ae1523659b96 | In this work we show that Deep Convolutional Neural Networks can outperform humans on the task of boundary detection, as measured on the standard Berkeley Segmentation Dataset. Our detector is fully integrated in the popular Caffe framework and processes a 320x420 image in less than a second. Our contributions consist firstly in combining a careful design of the loss for boundary detection training, a multi-resolution architecture and training with external data to improve the detection accuracy of the current state of the art, from an optimal dataset scale F-measure of 0.780 to 0.808 while human performance is at 0.803. We further improve performance to 0.813 by combining deep learning with grouping, integrating the Normalized Cuts technique within a deep network. We also examine the potential of our boundary detector in conjunction with the higher level tasks of object proposal generation and semantic segmentation for both tasks our detector yields clear improvements over state-of-the-art systems. |
7970532dc583554818922954ba518577e22a3cae | |
4869e04d79917d9b0b561a6d32103214d5ab2cfd | Effective cryptocurrency key management has become an urgent requirement for modern cryptocurrency. Although a large body of cryptocurrency wallet-management schemes has been proposed, they are mostly constructed for specific application scenarios and often suffer from weak security. In this paper, we propose a more effective, usable, and secure cryptocurrency wallet-management system based on semi-trusted social networks, therein allowing users to collaborate with involved parties to achieve some powerful functions and recovery under certain circumstances. Furthermore, we employ an identity-based hierarchical key-insulated encryption scheme to achieve time-sharing authorization and present a semi-trusted portable social-network-based wallet-management scheme that provides the features of security-enhanced storage, portable login on different devices, no-password authentication, flexible key delegation, and so on. The performance analysis shows that our proposed schemes require minimal additional overhead and have low time delays, making them sufficiently efficient for real-world deployment. |
82eb267b8e86be0b444e841b4b4ed4814b6f1942 | Understanding 3D object structure from a single image is an important but difficult task in computer vision, mostly due to the lack of 3D object annotations in real images. Previous work tackles this problem by either solving an optimization task given 2D keypoint positions, or training on synthetic data with ground truth 3D information. In this work, we propose 3D INterpreter Network (3D-INN), an endto-end framework which sequentially estimates 2D keypoint heatmaps and 3D object structure, trained on both real 2D-annotated images and synthetic 3D data. This is made possible mainly by two technical innovations. First, we propose a Projection Layer, which projects estimated 3D structure to 2D space, so that 3D-INN can be trained to predict 3D structural parameters supervised by 2D annotations on real images. Second, heatmaps of keypoints serve as an intermediate representation connecting real and synthetic data, enabling 3D-INN to benefit from the variation and abundance of synthetic 3D objects, without suffering from the difference between the statistics of real and synthesized images due to imperfect rendering. The network achieves state-of-the-art performance on both 2D keypoint estimation and 3D structure recovery. We also show that the recovered 3D information can be used in other vision applications, such as 3D rendering and image retrieval. |
04f818827a2ad16bf6d8585f45fba703c509c57b | We derive a second-order ordinary differential equation (ODE), which is the limit of Nesterov’s accelerated gradient method. This ODE exhibits approximate equivalence to Nesterov’s scheme and thus can serve as a tool for analysis. We show that the continuous time ODE allows for a better understanding of Nesterov’s scheme. As a byproduct, we obtain a family of schemes with similar convergence rates. The ODE interpretation also suggests restarting Nesterov’s scheme leading to an algorithm, which can be rigorously proven to converge at a linear rate whenever the objective is strongly convex. |
90140914a3a8d68bfe68f50821661f6af7bba5be | An effective defense-in-depth in cyber security applies multiple layers of defense throughout a system. The goalis to defend a system against cyber-attack using severalindependent methods. Therefore, a cyber-attack that is able to penetrate one layer of defense may be unsuccessful in other layers. Common layers of cyber defense include: attack avoidance, prevention, detection, survivability and recovery. It follows that in security-conscious organizations, the cyber security investment portfolio is divided into different layers of defense. For instance, a two-way division is agility and recovery. Cyber agility pursues attack avoidance techniques such that cyber-attacks are rendered as ineffective, whereas cyber recovery seeks to fight-through successful attacks. We show that even when the primary focus is on the agility of a system, recovery should be an essential point during implementation because the frequency of attacks will degrade the system and a quick and fast recovery is necessary. However, there is not yet an optimum mechanism to allocate limited cyber security resourcesinto the different layers. We propose an approach using theMarkov Decision Process (MDP) framework for resourcesallocation between the two end layers: agility and recovery. |
4fc2cc541fc8e85d5bfef6c6f8be97765462494b | This study applies the Unified Theory of Acceptance and Use of Technology (UTAUT) to the phenomenon of physician adoption of electronic medical records (EMR) technology. UTAUT integrates eight theories of individual acceptance into one comprehensive model designed to assist in understanding what factors either enable or hinder technology adoption and use. As such, it provides a useful lens through which to view what is currently taking place in the healthcare industry regarding EMR adoption. This is mutually beneficial to both the healthcare and MIS communities, as UTAUT offers valuable practical insight to the healthcare industry in explaining why EMR technology has not been more widely adopted as well as what prescriptions may facilitate future adoption, while offering the MIS community the opportunity to strengthen existing theory through an illustration of its application. |
09dc808c3249bbacb28a1e7b9c234cb58ad7dab4 | In this paper the evolution of silicon based automotive radar for the 76-81 GHz range is described. Starting with SiGe bare-die chips in 2009, today packaged MMICs are available for low-cost radar applications. Future SiGe BiCMOS technology will enable highly integrated single chip radars with superior performance at lower power consumption. This will pave the way for automotive radar safety for everyone and will be an important step towards autonomous driving. |
1bb6b92e874bf7cab8748a9acbb71a8dcda9ec65 | This article discusses the benefits and challenges associated with the design of multi-function scalable phased arrays at millimeter wave frequencies. First, applications for phased arrays with tens to hundreds of elements are discussed. Existing solutions for scaling silicon-based phased arrays from microwave to terahertz frequencies are reviewed. The challenges and tradeoffs associated with multiple integration options for W-band phased arrays are analyzed, with special consideration given to packaging and antenna performance. Finally, a solution based on SiGe ICs and organic packages for a 64-element dual-polarized 94 GHz phased array is described, along with associated measurement results. |
42cf4dd14f07cf34d5cad14fea5c691f31ab19ef | In this paper, a 77-GHz radar receiver is presented, which comes in a wafer level package and thus eliminates the need for wire bonding yielding significant cost reduction. The high integration level available in the productive Silicon-Germanium (SiGe) technology used in this paper allows for implementation of in-system monitoring of the receiver conversion parameters. This facilitates the realization of ISO 26262 compliant radar sensors for automotive safety applications. |
4a97f8d83c5cc523110beb11913ea5fb2a4271fe | A fully-integrated antenna-in-package (AiP) solution for W-band scalable phased-array systems is demonstrated. We present a fully operational compact W-band transceiver package with 64 dual-polarization antennas embedded in a multilayer organic substrate. This package has 12 metal layers, a size of 16.2 mm × 16.2 mm, and 292 ball-grid-array (BGA) pins with 0.4 mm pitch. Four silicon-germanium (SiGe) transceiver ICs are flip-chip attached to the package. Extensive full-wave electromagnetic simulation and radiation pattern measurements have been performed to optimize the antenna performance in the package environment, with excellent model-to-hardware correlation achieved. Enabled by detailed circuit-package co-design, a half-wavelength spacing, i.e., 1.6 mm at 94 GHz, is maintained between adjacent antenna elements to support array scalability at both the package and board level. Effective isotropic radiated power (EIRP) and radiation patterns are also measured to demonstrate the 64-element spatial power combining. |
76d7ddd030ef5ab4403a53caf2b46de89af71544 | A low-cost high-gain and broadband substrate integrated waveguide (SIW)-fed patch antenna array is demonstrated at the 60-GHz band. A single-layered SIW feeding network with wideband T-junctions and wideband high-gain cavity-backed patch antennas are employed to achieve high gain and wideband performance simultaneously. Although the proposed antenna array has a multilayered structure, it can be fabricated by conventional low-cost single-layered printed circuit board (PCB) technology and then realized by stacking and fixing all of single layers together. The simulated and measured impedance bandwidths of a 4 × 4 antenna array are 27.5% and 22.6% for 10 dB. The discrepancy between simulation and measurement is analyzed. A gain up to 19.6 dBi, and symmetrical unidirectional radiation patterns with low cross polarization are also achieved. With advantages of low fabrication cost and good performances, the proposed antenna array is a promising candidate for millimeter-wave wireless communication systems. |
ba2c1d97590651e152657eba5cb9e0a37d7d74d8 | In this paper, we present a systematic approach to the design, optimization and characterization of a broadband 5:1 bandwidth (0.8 to 4.0 GHz) antenna subarray. The array element is an optimized-taper antipodal Vivaldi slot with a bandwidth of 2.5:1. Two such elements of different sizes and with 0.4 GHz (10% of the highest frequency) overlapping bandwidths are arrayed in a nested lattice above a multilevel ground plane that shields the feeds and electronics. Return loss, radiation patterns, cross-polarization and mutual coupling are measured from 0.5–5.0 GHz. This array demonstrates E plane patterns with 50 and 45 3-dB beamwidths in the lower and upper frequency bands, respectively. The coupling between the elements is characterized for different relative antenna positions in all three dimensions. |
6b880d58b826074c49d0a38d0f9fb481e93c0acc | |
6a307cf51517f6c97d855639129afd002634ab92 | In this paper we present the novel SD-Map algorithm for exhaustive but efficient subgroup discovery. SD-Map guarantees to identify all interesting subgroup patterns contained in a data set, in contrast to heuristic or samplingbased methods. The SD-Map algorithm utilizes the well-known FP-growth method for mining association rules with adaptations for the subgroup discovery task. We show how SD-Map can handle missing values, and provide an experimental evaluation of the performance of the algorithm using synthetic data. |
1ea6b2f67a3a7f044209aae0d0fd1cb14a1e9e06 | Modeling the distribution of natural images is a landmark problem in unsupervised learning. This task requires an image model that is at once expressive, tractable and scalable. We present a deep neural network that sequentially predicts the pixels in an image along the two spatial dimensions. Our method models the discrete probability of the raw pixel values and encodes the complete set of dependencies in the image. Architectural novelties include fast twodimensional recurrent layers and an effective use of residual connections in deep recurrent networks. We achieve log-likelihood scores on natural images that are considerably better than the previous state of the art. Our main results also provide benchmarks on the diverse ImageNet dataset. Samples generated from the model appear crisp, varied and globally coherent. |
9871aa511ca7e3c61c083c327063442bc2c411bf | |
ba753286b9e2f32c5d5a7df08571262e257d2e53 | Generative Adversarial Nets [8] were recently introduced as a novel way to train generative models. In this work we introduce the conditional version of generative adversarial nets, which can be constructed by simply feeding the data, y, we wish to condition on to both the generator and discriminator. We show that this model can generate MNIST digits conditioned on class labels. We also illustrate how this model could be used to learn a multi-modal model, and provide preliminary examples of an application to image tagging in which we demonstrate how this approach can generate descriptive tags which are not part of training labels. |
2621c894885e3f42bca8bd2b11dab1637e697814 | OBJECTIVES
To achieve a better comprehension of the variability of perceptions, experiences and needs in terms of sexual and vaginal health in postmenopausal women (PMW) from four different European countries.
METHODS
An internet-based survey was conducted in Italy, Germany, Spain and the United Kingdom with a total surveyed population of 3768 PMW aged between 45 and 75 years.
RESULTS
The UK sample was significantly older, with almost a quarter of participants over 65 years of age, and had the highest proportion of women experiencing recent vulvar and vaginal atrophy (52.8%). The majority of Italian and Spanish participants were receiving VVA treatment, whereas in the UK only 28% of PMW were on medication. The most common menopausal symptom was vaginal/vulvar dryness, with almost 80% of participants reporting it in all the countries except the UK (48%). On the other hand, vaginal/vulvar irritation was more frequently reported in the UK (41%). The percentage of participants with a partner was lower in the UK (71%), as was the monthly rate of sexual activity (49%). In the UK, the proportion of participants who had seen a healthcare professional for gynaecological reasons in the last year was lower than in other countries (27% vs. ≥50%), as was the proportion who has discussed their VVA symptoms with them (45% vs. ∼67%). In this sense, UK PMW waited for a longer before asking for help (especially for pain with intercourse and dryness). The main issues relating to VVA treatment difficulties expressed by participants were administration route in the UK, efficacy in Germany, and side-effects in Italy.
CONCLUSIONS
Although all European women shared the same expectation of improving the quality of their sex lives, the opportunity for that varied among different countries in relation to the healthcare system and to the effective communication achieved with healthcare professionals when managing VVA. |
46a26a258bbd087240332376a6ead5f1f7cf497a | JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. This article concerns theories about why and how information technology affects organizational life. Good theory guides research, which, when applied, increases the likelihood that information technology will be employed with desirable consequences for users, organizations, and other interested parties. But what is a good theory? Theories are often evaluated in terms of their content-the specific concepts used and the human values served. This article examines theories in terms of their structures-theorists' assumptions about the nature and direction of causal influence. Three dimensions of causal structure are considered-causal agency, logical structure, and level of analysis. Causal agency refers to beliefs about the nature of causality: whether external forces cause change, whether people act purposefully to accomplish intended objectives, or whether changes emerge unpredictably from the interaction of people and events. Logical structure refers to the temporal aspect of theory-static versus dynamic-and to the logical relationships between the "causes" and the outcomes. Level of analysis refers to the entities about which the theory poses concepts and relationships-individuals, groups, organizations, and society. While there are many possible structures for good theory about the role of information technology in organizational change, only a few of these structures can be seen in current theorizing. Increased awareness of the options, open discussion of their advantages and disadvantages , and explicit characterization of future theoretical statements in terms of the dimensions and categories discussed here should, we believe, promote the development of better theory. |
21a39ca716c37ccc133ff96c31cc71565d0c968e | We present a novel representation of maps between pairs of shapes that allows for efficient inference and manipulation. Key to our approach is a generalization of the notion of map that puts in correspondence real-valued functions rather than points on the shapes. By choosing a multi-scale basis for the function space on each shape, such as the eigenfunctions of its Laplace-Beltrami operator, we obtain a representation of a map that is very compact, yet fully suitable for global inference. Perhaps more remarkably, most natural constraints on a map, such as descriptor preservation, landmark correspondences, part preservation and operator commutativity become linear in this formulation. Moreover, the representation naturally supports certain algebraic operations such as map sum, difference and composition, and enables a number of applications, such as function or annotation transfer without establishing point-to-point correspondences. We exploit these properties to devise an efficient shape matching method, at the core of which is a single linear solve. The new method achieves state-of-the-art results on an isometric shape matching benchmark. We also show how this representation can be used to improve the quality of maps produced by existing shape matching methods, and illustrate its usefulness in segmentation transfer and joint analysis of shape collections. |
1bdf90cc38336e6e627150e006fd58bcba201792 | Body Sensor Networks (BSNs) are conveying notable attention due to their capabilities in supporting humans in their daily life. In particular, real-time and noninvasive monitoring of assisted livings is having great potential in many application domains, such as health care, sport/fitness, e-entertainment, social interaction and e-factory. And the basic as well as crucial feature characterizing such systems is the ability of detecting human actions and behaviors. In this paper, a novel approach for human posture recognition is proposed. Our BSN system relies on an information fusion method based on the D-S Evidence Theory, which is applied on the accelerometer data coming from multiple wearable sensors. Experimental results demonstrate that the developed prototype system is able to achieve a recognition accuracy between 98.5% and 100% for basic postures (standing, sitting, lying, squatting). |
0ef98882d8a7356c7cf7ac715bef84656a0632d4 | Nested dichotomies are a standard statistical technique for tackling certain polytomous classification problems with logistic regression. They can be represented as binary trees that recursively split a multi-class classification task into a system of dichotomies and provide a statistically sound way of applying two-class learning algorithms to multi-class problems (assuming these algorithms generate class probability estimates). However, there are usually many candidate trees for a given problem and in the standard approach the choice of a particular tree is based on domain knowledge that may not be available in practice. An alternative is to treat every system of nested dichotomies as equally likely and to form an ensemble classifier based on this assumption. We show that this approach produces more accurate classifications than applying C4.5 and logistic regression directly to multi-class problems. Our results also show that ensembles of nested dichotomies produce more accurate classifiers than pairwise classification if both techniques are used with C4.5, and comparable results for logistic regression. Compared to error-correcting output codes, they are preferable if logistic regression is used, and comparable in the case of C4.5. An additional benefit is that they generate class probability estimates. Consequently they appear to be a good general-purpose method for applying binary classifiers to multi-class problems. |
3d0f7539d17816a5ba1925f10ed5c274c9d570b5 | The ability to predict performance of students is very crucial in our present education system. We can use data mining concepts for this purpose. ID3 algorithm is one of the famous algorithms present today to generate decision trees. But this algorithm has a shortcoming that it is inclined to attributes with many values. So , this research aims to overcome this shortcoming of the algorithm by using gain ratio(instead of information gain) as well as by giving weights to each attribute at every decision making point. Several other algorithms like J48 and Naive Bayes classification algorithm are also applied on the dataset. The WEKA tool was used for the analysis of J48 and Naive Bayes algorithms. The results are compared and presented. The dataset used in our study is taken from the School of Computing Sciences and Engineering (SCSE), VIT University. Keyworddata mining, educational data mining (EDM), decision tree, gain ratio, weighted ID3 |
2cef1515ac1d98f1dcfdc9b84c1a3b6f3758b9d1 | Human analysis in images and video is a hard problem due to the large variation in human pose, clothing, camera view-points, lighting and other factors. While the explicit modeling of this variability is difficult, the huge amount of available person images motivates for the implicit, data-driven approach to human analysis. In this work we aim to explore this approach using the large amount of images spanning a subspace of human appearance. We model this subspace by connecting images into a graph and propagating information through such a graph using a discriminatively-trained graphical model. We particularly address the problems of human pose estimation and action recognition and demonstrate how image graphs help solving these problems jointly. We report results on still images with human actions from the KTH dataset. |
033e667eec844477bdfcd0e4b3d5c81b7598af39 | Helical antennas have been known for a long time, but the literature is overwhelmed with controversial information about their performance. We have systematically investigated helical antennas located above an infinite ground plane and obtained design curves. We have also observed that the shape and size of the ground conductor have influence on the helical antenna performance. By optimizing the dimensions of ground conductors that have the form of a cup and a cone, we have significantly increased the antenna gain. Simultaneous optimization of the helix and the ground conductor is under way to improve the antenna performance. |
91e4456442a7bce77cdfd24bc8a2fb6a4a3a3b6f | Introduction Statistical language models estimate the probability for a given sequence of words. Given a sentence s with n words such as s = (w1,w2 . . .wn), the language model assigns P(s). Statistical language models assess good word sequence estimations based on the sequence probability estimation. Building robust, fast and accurate language models is one of the main factors in the success of building systems such as machine translation systems and automatic speech recognition systems. Statistical language models can be estimated based on various approaches. Classical language models are estimated based on n-gram word sequences such as P(s) = P(w1,w2 . . .wn) = ∏n i=1 P(wi|wi−1) and can be approximated based on the Markov concept for shorter contexts (for example, bigram if n = 2 or trigram if n = 3 and so on). Recent researchers have applied neural networks different architectures to build and estimate language models. The classical feed-forward neural network-based language models have been continuously reporting good results among the traditional n-gram language modeling techniques [1]. It Abstract |
f713a480be9d569803349338a1c2db8cabb0165c | A key challenge in complex visuomotor control is learning abstract representations that are effective for specifying goals, planning, and generalization. To this end, we introduce universal planning networks (UPN). UPNs embed differentiable planning within a goal-directed policy. This planning computation unrolls a forward model in a latent space and infers an optimal action plan through gradient descent trajectory optimization. The plan-by-gradient-descent process and its underlying representations are learned end-to-end to directly optimize a supervised imitation learning objective. We find that the representations learned are not only effective for goal-directed visual imitation via gradient-based trajectory optimization, but can also provide a metric for specifying goals using images. The learned representations can be leveraged to specify distance-based rewards to reach new target states for model-free reinforcement learning, resulting in substantially more effective learning when solving new tasks described via imagebased goals. Visit https://sites.google. com/view/upn-public/home for video highlights. |
c2c8292dad37adcdda3b0aba5db5e33e3222832e | |
5facbc5bd594f4faa0524d871d49ba6a6e956e17 | We propose a method based on sparse representation (SR) to cluster data drawn from multiple low-dimensional linear or affine subspaces embedded in a high-dimensional space. Our method is based on the fact that each point in a union of subspaces has a SR with respect to a dictionary formed by all other data points. In general, finding such a SR is NP hard. Our key contribution is to show that, under mild assumptions, the SR can be obtained `exactly' by using l1 optimization. The segmentation of the data is obtained by applying spectral clustering to a similarity matrix built from this SR. Our method can handle noise, outliers as well as missing data. We apply our subspace clustering algorithm to the problem of segmenting multiple motions in video. Experiments on 167 video sequences show that our approach significantly outperforms state-of-the-art methods. |
b98063ae46f00587a63547ce753c515766db4fbb | Microstrip patch antennas offer the advantages of thin profile, light weight, low cost, ease of fabrication and compatibility with integrated circuitry, so the antennas are widely used to satisfy demands for polarization diversity and dual-frequency. This paper presents a coaxilly-fed single-layer compact dual band (Cand X-band) microstrip patch antenna for achieving dual-polarized radiation suitable for applications in airborne synthetic aperture radar (SAR) systems. The designed antenna consists of three rectangular patches which are overlapped along their diagonals. Full-wave electromagnetic simulations are performed to accurately predict the frequency response of the antenna. The fabricated antenna achieves an impedance bandwidth of 154 MHz (f0 = 6.83 GHz) and 209 MHz (f0 = 9.73 GHz) for VSWR < 2. Simultaneous use of both frequencies should drastically improve data collection and knowledge of the targets in the SAR system. |
358d2a02275109c250f74f8af150d42eb75f7b5f | Public Key Cryptography (PKC) has been the enabling technology underlying many security services and protocols in traditional networks such as the Internet. In the context of wireless sensor networks, elliptic curve cryptography (ECC), one of the most efficient types of PKC, is being investigated to provide PKC supportin sensor network applications so that the existing PKC-based solutions can be exploited. This paper presents the design, implementation, and evaluation of TinyECC, a configurable library for ECC operations in wireless sensor networks. The primary objective of TinyECC is to provide a ready-to-use, publicly available software package for ECC-based PKC operations that can be flexibly configured and integrated into sensor network applications. TinyECC provides a number of optimization switches, which can turn specific optimizations on or off based on developers' needs. Different combinations of the optimizations have different execution time andresource consumptions, giving developers great flexibility in integrating TinyECC into sensor network applications. This paperalso reports the experimental evaluation of TinyECC on several common sensor platforms, including MICAz, Tmote Sky, and Imote2. The evaluation results show the impacts of individual optimizations on the execution time and resource consumptions, and give the most computationally efficient and the most storage efficient configuration of TinyECC. |
d1d044d0a94f6d8ad1a00918abededbd2a41f1bd | * Corresponding author: Ch.Sandeep Kumar Subudhi Abstract Aim of our work is to monitor the human body temperature, blood pressure (BP), Pulse Rate and ECG and tracking the patient location. The human body temperature, BP, Pulse Rate and ECG are detected in the working environment; this can be sensed by using respective sensors. The sensed information is send to the PIC16F877 microcontroller through signal conditioning circuit in the patient unit. A desired amount of sensor value is set and if it is exceeded preliminary steps should be taken by the indicating by buzzer.The sensor information will be transmitted from the patient unit to the main controller unit with the help of Zigbee communication system which is connected with the microcontrollers in the both units. The main controller unit will send those sensed data as well as the location of that patient by the help of GPS Module to the observer/doctor. The observer/doctor can receive the SMS sent by GSM module and further decision can be taken. The message is sent to a mobile phone using Global system mobile (GSM) Modem. MAX232 was a driver between microcontroller and modem. |
f980462aa340a1beed2f040db9b54d7f22378a58 | Stereo vision is an active research topic in computer vision. Point Grey® Bumblebee® and digital single-lens reflex camera (DSLR) are normally found in the stereo vision research, they are robust but expensive. Open source electronic prototyping platforms such as Arduino and Raspberry Pi are interesting products, which allows students or researchers to custom made inexpensive experimental equipment for their research projects. This paper describes the intelligent stereo camera mobile platform developed in our research using Pi and camera modules and presents the concept of using inexpensive open source parts for robotic stereo vision research work in details. |
11db74171df92a50d64bd88d569454415878c63a | We present a parser for Abstract Meaning Representation (AMR). We treat Englishto-AMR conversion within the framework of string-to-tree, syntax-based machine translation (SBMT). To make this work, we transform the AMR structure into a form suitable for the mechanics of SBMT and useful for modeling. We introduce an AMR-specific language model and add data and features drawn from semantic resources. Our resulting AMR parser improves upon state-of-the-art results by 7 Smatch points. |
f99f46f1b90be4582946951f4f12d8b9209354b2 | An antenna consisting of 4 element subarrays of strongly coupled E-shaped patches placed conformally on a metal cylinder is proposed for use in wireless local area networks (WLAN). The use of this special array configuration in the cylindrical case allows reaching a 8.3 % bandwidth and an omnidirectional radiation pattern in the horizontal plane with only 2 subarrays. CST Microwave Studio is used for validation purposes before manufacturing. To the knowledge of the authors, it is the first time that this recently introduced concept of strongly coupled E-shaped patches is used in a conformal antenna. |
b72a44e0f509a91a02c4d5b2df2a95fb1b3ee392 | This article introduces the Follow-Me Cloud concept and proposes its framework. The proposed framework is aimed at smooth migration of all or only a required portion of an ongoing IP service between a data center and user equipment of a 3GPP mobile network to another optimal DC with no service disruption. The service migration and continuity is supported by replacing IP addressing with service identification. Indeed, an FMC service/application is identified, upon establishment, by a session/service ID, dynamically changing along with the service being delivered over the session; it consists of a unique identifier of UE within the 3GPP mobile network, an identifier of the cloud service, and dynamically changing characteristics of the cloud service. Service migration in FMC is triggered by change in the IP address of the UE due to a change of data anchor gateway in the mobile network, in turn due to UE mobility and/or for load balancing. An optimal DC is then selected based on the features of the new data anchor gateway. Smooth service migration and continuity are supported thanks to logic installed at UE and DCs that maps features of IP flows to the session/service ID. |
11190a466d1085c09a11e52cc63f112280ddce74 | A visual attention system, inspired by the behavior and the neuronal architecture of the early primate visual system, is presented. Multiscale image features are combined into a single topographical saliency map. A dynamical neural network then selects attended locations in order of decreasing saliency. The system breaks down the complex problem of scene understanding by rapidly selecting, in a computationally efficient manner, conspicuous locations to be analyzed in detail. |
5b2bc4aaa63412ff1745a79d2f322b5ff67d0f9c | |
b2db00f73fc6b97ebe12e97cfdaefbb2fefc253b | One of the most difficult problems in the design of an anomaly b sed intrusion detection system (IDS) that uses clustering is th at of labelling the obtained clusters, i.e. determining which of them correspond t ”good” behaviour on the network/host and which to ”bad” behaviour. In this pap er, a new clusters’ labelling strategy, which makes use of a clustering quality index is proposed for application in such an IDS. The aim of the new labelling algor ithm is to detect compact clusters containing very similar vectors and these are highly likely to be attack vectors. Two clustering quality indexes have been te sted and compared: the Silhouette index and the Davies-Bouldin index. Experiment al results comparing the effectiveness of a multiple classifier IDS with the two in dexes implemented show that the system using the Silhouette index produces sli ghtly more accurate results than the system that uses the Davies-Bouldin index. However, the computation of the Davies-Bouldin index is much less complex th an the computation of the Silhouette index, which is a very important advantage regarding eventual real-time operation of an IDS that employs clustering. |
ee35c52c22fadf92277c308263be6288249a6327 | Previous studies on Vivaldi wideband phased arrays are mainly focused on Two-dimensional scanning phased arrays. A miniaturized balanced antipodal Vivaldi antenna (BAVA) is presented in this paper. A novel vertical parasitic metal strip loading is employed in the dual-polarized linear phased arrays, it to make a compact structure. With the arc-shaped slots and metal strip loads, the radiation performance in lower operating band can be greatly enhanced. The proposed antenna is simulated in the infinite condition through periodic boundary with the size of 100mm (length) *100mm (width) *125mm (depth). The antenna achieves an impedance bandwidth when scanning to ±50° for VSWR≤3 achieves 4:1 (0.5GHz-2GHz) bandwidth and 5:1 (0.4GHz-2GHz) bandwidth in the vertical polarization and horizontal polarization, respectively, and the isolation is less than -18dB over the operating frequency range. |
62198d1c71b827f0674da3d4d7ebf718702713eb | Ultra-wideband (UWB) pulse Doppler radars can be used for noncontact vital signs monitoring of more than one subject. However, their detected signals typically have low signal-to-noise ratio (SNR) causing significant heart rate (HR) detection errors, as the spurious harmonics of respiration signals and mixed products of respiration and heartbeat signals (that can be relatively higher than heartbeat signals) corrupt conventional fast Fourier transform spectrograms. In this paper, we extend the complex signal demodulation (CSD) and arctangent demodulation (AD) techniques previously used for accurately detecting the phase variations of reflected signals of continuous wave radars to UWB pulse radars as well. These detection techniques reduce the impact of the interfering harmonic signals, thus improving the SNR of the detected vital sign signals. To further enhance the accuracy of the HR estimation, a recently developed state-space method has been successfully combined with CSD and AD techniques and over 10 dB improvements in SNR is demonstrated. The implementation of these various detection techniques has been experimentally investigated and full error and SNR analysis of the HR detection are presented. |
962f9eec6cce6253a2242fad2461746c9ccd2a0f | A novel via-less coplanar waveguide (CPW) to microstrip transition is discussed and design rules based on simulations and experimental results are presented. This transition demonstrates a maximum insertion loss of 1 dB over the frequency range from 10 GHz to 40 GHz with a value of 0.4 dB at 20 GHz. This transition could find a variety of applications due to its compatibility with RF systems-on-a chip, low loss performance, low cost and its ease of fabrication. |
987878e987c493de52efa5b96db6558a65374585 | A metamaterial balun that converts a single-ended input to a differential output over a large bandwidth is presented. The device also exhibits excellent return loss, isolation, and through characteristics over the same frequency band. The balun comprises a Wilkinson divider, followed by a +90/spl deg/ negative-refractive-index (NRI) metamaterial (MM) phase-shifting line along the top branch, and a -90/spl deg/ MM phase-shifting line along the bottom branch. Utilizing MM lines for both the +90/spl deg/ and -90/spl deg/ branches allows the slopes of their phase responses to be matched, resulting in a broadband differential output signal. The theoretical performance of the balun is verified through circuit simulations and measurements of a fabricated prototype at 1.5 GHz. The MM balun exhibits a measured differential output phase bandwidth (180/spl deg//spl plusmn/10/spl deg/) of 1.16 GHz (77%), from 1.17 to 2.33 GHz. The measured isolation and return loss for all three ports remain below -10 dB over a bandwidth in excess of 2 GHz, while the output quantities |S/sub 21/| and |S/sub 31/| remain above -4 dB from 0.5 to 2.5 GHz. |
39da8410e503738eb19cd5d2f3e154e7e2a9971b | In this paper, we present an image-analysis based approach to calorie content estimation for dietary assessment. We make use of daily food images captured and stored by multiple users in a public Web service called Food Log. The images are taken without any control or markers. We build a dictionary dataset of 6512 images contained in Food Log the calorie content of which have been estimated by experts in nutrition. An image is compared to the ground truth data from the point of views of multiple image features such as color histograms, color correlograms and SURF fetures, and the ground truth images are ranked by similarities. Finally, calorie content of the input food image is computed by linear estimation using the top n ranked calories in multiple features. The distribution of the estimation shows that 79% of the estimations are correct within ±40% error and 35% correct within ±20% error. |
164bd38ee518e8191fc3fb27b23a10c2668f76d9 | Fig. 1. Given a natural image, the proposed approach can hallucinate different versions of the same scene in a wide range of conditions, e.g.night, sunset, winter, spring, rain, fog or even a combination of those. First, our method utilizes a single generator network to imagine the scene with respect to its semantic layout and the desired set of attributes. Then, it directly transfers the look from the hallucinated output to the input image, without a need to have access to a reference style image. |
08d94e04e4be38ef06bd0a3294af3936a27db529 | The log periodic antenna and Yagi-Uda antenna are used in the applications where very high directivity is required. They also give very high gain in the range of 17-20dBi. This paper presents a review on various configurations of log periodic and Yagi antennas, their advantages and problems. One problem encountered with Yagi-Uda antenna is relatively less bandwidth. This problem is solved by log periodic antenna which can operate over high bandwidth and providing high gain at the same time. In this paper, a review of various techniques to realize printed Yagi-Uda and log periodic antenna is discussed. They are realized by using different feeding techniques like microstrip feeding, co-axial feeding etc. They are also realized by using modifying the shape of directors and reflectors. The high bandwidth (log periodic antenna) has also been realized by increasing the number of reflectors. |
2e0b1facffd6e9a0b1bc6b87d1dab0874846fee0 | To sustain perpetual operation, systems that harvest environmental energy must carefully regulate their usage to satisfy their demand. Regulating energy usage is challenging if a system's demands are not elastic and its hardware components are not energy-proportional, since it cannot precisely scale its usage to match its supply. Instead, the system must choose when to satisfy its energy demands based on its current energy reserves and predictions of its future energy supply. In this paper, we explore the use of weather forecasts to improve a system's ability to satisfy demand by improving its predictions. We analyze weather forecast, observational, and energy harvesting data to formulate a model that translates a weather forecast to a wind or solar energy harvesting prediction, and quantify its accuracy. We evaluate our model for both energy sources in the context of two different energy harvesting sensor systems with inelastic demands: a sensor testbed that leases sensors to external users and a lexicographically fair sensor network that maintains steady node sensing rates. We show that using weather forecasts in both wind- and solar-powered sensor systems increases each system's ability to satisfy its demands compared with existing prediction strategies. |
87a851593965e371184cf3c3df5630f8d66bf49b | The rapidly growing VLSI technology could reduce the cost of the embedded systems considerably which has motivated a new era of low cost personal health care monitoring systems which meeting common man requirements. Such systems are dedicated to one person and provide the diagnostic information through remotely. This paper proposes a PSoC microcontroller and GSM modules we are eliminating the cables attached to patient. The patient has a freedom of doing daily activities and still be under continuous monitoring (very suitable for old age people). PSoC has inbuilt ADC and Programmable Gain amplifier which enabled single chip implementation. The hardware complexity is also simple and reduces cost. The basic principle of this system is to read the bio medical signals from the biomedical sensor modules and perform the tasks of data conversion, sending SMS using GSM, as well as providing the ability of simple preprocessing such as waveform averaging or rectification. GSM based modem interface through AT commands to PSoC. The Microcontroller will be able to receive and send text messages Heart beat is sensed by using a high intensity type LED and LDR. The finger is placed between the LED and LDR. The skin may be illuminated with visible (red) using transmitted or reflected light for detection. The concept of this paper is builds upon the integration of wireless communications into medical applications to revolutionize personal healthcare. Under this concept, patients are no longer bound to a specific healthcare location where they are monitored by medical instruments—wireless communications would not only provide them with safe and accurate monitoring but also the freedom of movement. Remote patient monitoring will not only redefine hospital care but also work, home, and recreational activities. Imagine the home of the future: elderly individuals no longer have to travel to the hospital, a mother stuck at work will be able to receive email updates of her sick child's temperature, a doctor on the go will have the ability to turn on his cell phone to check the status of his patients. The objective of this project is to build a wireless heart beat monitoring system using GSM Technology, which could potentially be an integral part of a suite of personal healthcare appliances for a large-scale remote patient monitoring system. In this paper a low cost Digital Heart rate meter using PSoC Microcontroller is designed, and also this unit is connecting to GSM Modem for send … |
0558c94a094158ecd64f0d5014d3d9668054fb97 | We present Resilient Distributed Datasets (RDDs), a distributed memory abstraction that allows programmers to perform in-memory computations on large clusters while retaining the fault tolerance of data flow models like MapReduce. RDDs are motivated by two types of applications that current data flow systems handle inefficiently: iterative algorithms, which are common in graph applications and machine learning, and interactive data mining tools. In both cases, keeping data in memory can improve performance by an order of magnitude. To achieve fault tolerance efficiently, RDDs provide a highly restricted form of shared memory: they are read-only datasets that can only be constructed through bulk operations on other RDDs. However, we show that RDDs are expressive enough to capture a wide class of computations, including MapReduce and specialized programming models for iterative jobs such as Pregel. Our implementation of RDDs can outperform Hadoop by 20× for iterative jobs and can be used interactively to search a 1 TB dataset with latencies of 5–7 seconds. |
c77be34db96695159244723fe9ffa4a88dc4a36d | Understanding and estimating satisfaction with search engines is an important aspect of evaluating retrieval performance. Research to date has modeled and predicted search satisfaction on a binary scale, i.e., the searchers are either satisfied or dissatisfied with their search outcome. However, users' search experience is a complex construct and there are different degrees of satisfaction. As such, binary classification of satisfaction may be limiting. To the best of our knowledge, we are the first to study the problem of understanding and predicting graded (multi-level) search satisfaction. We ex-amine sessions mined from search engine logs, where searcher satisfaction was also assessed on multi-point scale by human annotators. Leveraging these search log data, we observe rich and non-monotonous changes in search behavior in sessions with different degrees of satisfaction. The findings suggest that we should predict finer-grained satisfaction levels. To address this issue, we model search satisfaction using features indicating search outcome, search effort, and changes in both outcome and effort during a session. We show that our approach can predict subtle changes in search satisfaction more accurately than state-of-the-art methods, affording greater insight into search satisfaction. The strong performance of our models has implications for search providers seeking to accu-rately measure satisfaction with their services. |
d7d9d643a378b6fd69fff63d113f4eae1983adc8 | |
5e38e39dee575b07fb751bff5b05064b6942e2d3 | Students' dropout rate is a key metric in online and open distance learning courses. We propose a time-series classification method to construct data based on students' behaviour and activities on a number of online distance learning modules. Further, we propose a dropout prediction model based on the time series forest (TSF) classification algorithm. The proposed predictive model is based on interaction data and is independent of learning objectives and subject domains. The model enables prediction of dropout rates without the requirement for pedagogical experts. Results show that the prediction accuracy on two selected datasets increases as the portion of data used in the model grows. However, a reasonable prediction accuracy of 0.84 is possible with only 5% of the dataset processed. As a result, early prediction can help instructors design interventions to encourage course completion before a student falls too far behind. |
625fb5f70406ac8dbd954d1105bd8e725d9254d9 | Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. |
b959164d1efca4b73986ba5d21e664aadbbc0457 | A quantitative and practical Bayesian framework is described for learning of mappings in feedforward networks. The framework makes possible (1) objective comparisons between solutions using alternative network architectures, (2) objective stopping rules for network pruning or growing procedures, (3) objective choice of magnitude and type of weight decay terms or additive regularizers (for penalizing large weights, etc.), (4) a measure of the effective number of well-determined parameters in a model, (5) quantified estimates of the error bars on network parameters and on network output, and (6) objective comparisons with alternative learning and interpolation models such as splines and radial basis functions. The Bayesian "evidence" automatically embodies "Occam's razor," penalizing overflexible and overcomplex models. The Bayesian approach helps detect poor underlying assumptions in learning models. For learning models well matched to a problem, a good correlation between generalization ability and the Bayesian evidence is obtained. |
27578f85abf2cc167e855411bbfa972d7d26ec0f | An FET-sized 1-18 GHz monolithic active magic T (1 W hybrid) is proposed. It unifies two different dividers, electrically isolated from each other, in a novel GaAs FET electrode configuration, viz. the LUFET concept. Its characteristics and experiment results are presented. Applications of the magic T to miniature wide-band RF signal processing. . . |
25642be46de0f2e74e0da81a14646f8bfcc9000a | Image classification is a critical task for both humans and computers. One of the challenges lies in the large scale of the semantic space. In particular, humans can recognize tens of thousands of object classes and scenes. No computer vision algorithm today has been tested at this scale. This paper presents a study of large scale categorization including a series of challenging experiments on classification with more than 10, 000 image classes. We find that a) computational issues become crucial in algorithm design; b) conventional wisdom from a couple of hundred image categories on relative performance of different classifiers does not necessarily hold when the number of categories increases; c) there is a surprisingly strong relationship between the structure of WordNet (developed for studying language) and the difficulty of visual categorization; d) classification can be improved by exploiting the semantic hierarchy. Toward the future goal of developing automatic vision algorithms to recognize tens of thousands or even millions of image categories, we make a series of observations and arguments about dataset scale, category density, and image hierarchy. |
b8b17b2dd75749ecd68eb1ff1d2cea1703660a18 | For the past several years, the Defense Advanced Research Projects Agency (DARPA) has been pioneering the development of the first ever real-time knowledge-aided (KA) adaptive radar architecture. The impetus for this program is the ever increasingly complex missions and operational environments encountered by modern radars and the inability of traditional adaptation methods to address rapidly varying interference environments. The DARPA KA sensor signal processing and expert reasoning (KASSPER) program has as its goal the demonstration of a high performance embedded computing (HPEC) architecture capable of integrating high-fidelity environmental knowledge (i.e., priors) into the most computationally demanding subsystem of a modern radar: the adaptive space-time beamformer. This is no mean feat as environmental knowledge is a memory quantity that is inherently difficult (if not impossible) to access at the rates required to meet radar front-end throughput requirements. In this article, we will provide an overview of the KASSPER program highlighting both the benefits of KA adaptive radar, key algorithmic concepts, and the breakthrough look-ahead radar scheduling approach that is the keystone to the KASSPER HPEC architecture. |
0f4f5ba66a0b666c512c4f120c521cecc89e013f | The current evolution of the traditional medical model toward the participatory medicine can be boosted by the Internet of Things (IoT) paradigm involving sensors (environmental, wearable, and implanted) spread inside domestic environments with the purpose to monitor the user's health and activate remote assistance. RF identification (RFID) technology is now mature to provide part of the IoT physical layer for the personal healthcare in smart environments through low-cost, energy-autonomous, and disposable sensors. It is here presented a survey on the state-of-the-art of RFID for application to body centric systems and for gathering information (temperature, humidity, and other gases) about the user's living environment. Many available options are described up to the application level with some examples of RFID systems able to collect and process multichannel data about the human behavior in compliance with the power exposure and sanitary regulations. Open challenges and possible new research trends are finally discussed. |
f9cd7da733c5b5b54a4bbd35f67a913c05df83ea | Jammer and interference are sources of errors in positions estimated by GNSS receivers. The interfering signals reduce signal-to-noise ratio and cause receiver failure to correctly detect satellite signals. Because of the robustness of beamforming techniques to jamming and multipath mitigation by placing nulls in direction of interference signals, an antenna array with a set of multi-channel receivers can be used to improve GNSS signal reception. Spatial reference beam forming uses the information in the Direction Of Arrival (DOA) of desired and interference signals for this purpose. However, using a multi-channel receiver is not applicable in many applications for estimating the Angle Of Arrival (AOA) of the signal (hardware limitations or portability issues). This paper proposes a new method for DOA estimation of jammer and interference signals based on a synthetic antenna array. In this case, the motion of a single antenna can be used to estimate the AOA of the interfering signals. |
77f14f3b5f094bf2e785fae772846116da18fa48 | Similes are easily obtained from web-driven and casebased reasoning approaches. Still, generating thoughtful figurative descriptions with meaningful relation to narrative context and author style has not yet been fully explored. In this paper, the author prepares the foundation for a computational model which can achieve this level of aesthetic complexity. This paper also introduces and evaluates a possible architecture for generating and ranking figurative comparisons on par with humans: the |
8759c972d89e1bf6deeab780aa2f8e21140c953b | A sidelobe suppression method for a series slot array antenna which radiates 45° -inclined linear polarization is proposed. Axial displacements are employed to create arbitrary excitation coefficients for individual centered-inclined radiating slots along the center line of a broad wall. To verify the proposed design method, we design two types of center-fed linear slot array antennas with a Dolph-Chebyshev distribution for -20 dB and -26 dB sidelobe levels (SLLs) in the Ka band. Furthermore, a cross-validation process involving an equivalent circuit model analysis and electromagnetic full-wave simulation using CST MWS is utilized. The entire structure of the proposed series slot array antenna is fabricated on printed circuit boards (PCBs), including drilling and chemical etching, to secure advantages of miniaturization and cost reduction. The measured realized gains are 15.17 and 15.95 dBi and SLLs are -18.7 and -22.5 dB respectively for two types of fabricated antennas. It demonstrates the validity of the proposed sidelobe suppression method. |
b790ed1ac0b3451ff7522b3b2b9cda6ca3e28670 | |
dde75bf73fa0037cbf586d89d698825bae2a4669 | We propose a compact conformal wearable antenna that operates in the 2.36-2.4 GHz medical body-area network band. The antenna is enabled by placing a highly truncated metasurface, consisting of only a two by two array of I-shaped elements, underneath a planar monopole. In contrast to previously reported artificial magnetic conducting ground plane backed antenna designs, here the metasurface acts not only as a ground plane for isolation, but also as the main radiator. An antenna prototype was fabricated and tested, showing a strong agreement between simulation and measurement. Comparing to previously proposed wearable antennas, the demonstrated antenna has a compact form factor of 0.5 λ0 ×0.3 λ0 ×0.028 λ0, all while achieving a 5.5% impedance bandwidth, a gain of 6.2 dBi, and a front-to-back ratio higher than 23 dB. Further numerical and experimental investigations reveal that the performance of the antenna is extraordinarily robust to both structural deformation and human body loading, far superior to both planar monopoles and microstrip patch antennas. Additionally, the introduced metal backed metasurface enables a 95.3% reduction in the specific absorption rate, making such an antenna a prime candidate for incorporation into various wearable devices. |
a3442990ba9059708272580f0eef5f84f8605427 | Network Functions Virtualization (NFV) has recently emerged as one of the major technological driving forces that significantly accelerate the evolution of today's computer and communication networks. Despite the advantages of NFV, e.g., saving investment cost, optimizing resource consumption, improving operational efficiency, simplifying network service lifecycle management, lots of novel security threats and vulnerabilities will be introduced, thereby impeding its further development and deployment in practice. In this paper, we briefly report our threat analysis in the context of NFV, and identify the corresponding security requirements. The purpose is to establish a comprehensive threat taxonomy and provide a guideline to develop effective security countermeasures. Furthermore, a conceptual design framework for NFV based security management and service orchestration is presented, with an objective to dynamically and adaptively deploy and manage security functions on the demands of users and customers. A use case about NFV based access control is also developed, illustrating the feasibility and advantages of implementing NFV based security management and orchestration. |
5e6f43abce81b991a0860f28f0915b2333b280c5 | Users are constantly involved in a multitude of activities in ever-changing context. Analyzing activities in contextrich environments has become a great challenge in contextawareness research. Traditional methods for activity recognition, such as classification, cannot cope with the variety and dynamicity of context and activities. In this paper, we propose an activity recognition approach that incorporates unsupervised learning. We analyze the feasibility of applying subspace clustering—a specific type of unsupervised learning— to high-dimensional, heterogeneous sensory input. Then we present the correspondence between clustering output and classification input. This approach has the potential to discover implicit, evolving activities, and can provide valuable assistance to traditional classification based methods. As sensors become prevalent means in context detection and information channels proliferate to make context sharing easier, it is increasingly challenging to interpret context and analyze its effects on the activities (Lim and Dey 2010). We argue that applying traditional approaches to activity recognition may become more and more difficult to apply in context and activity-rich environments. In the literature, context attributes used for learning activities are chosen by either empirical assumption or dimension reduction to render a small set of features (Krause, Smailagic, and Siewiorek 2006). These approaches are infeasible in face of a broad spectrum of context information. The most significant drawback is that they fail to acknowledge the large variety of features needed to describe different activities. For activity recognition, most previous works applied supervised learning approaches that aimed at predicting activities among a set of known classes(Ferscha et al. 2004). These approaches, however, are also challenged when coping with new and fast evolving activities. Unsupervised learning, particularly clustering, has been highly successful for revealing implicit relationships and regularities in large data sets. Intuitively, we can envisage an activity recognition approach that applies clustering to context history. The clusters, representing frequent context patterns, can suggest activities and their contextual conditions. The results can be used independently for analyzing and interpreting activities. Furthermore, clusters can reveal the scopes and conditions Copyright c © 2011, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. of activities and interactions. This information is valuable when determining the scopes of information sharing in pervasive environments. Although clustering is a promising approach in discovering associations within context, the feasibility of traditional clustering is questionable in dealing with high dimensionality and heterogeneity of context data. In this paper, we will first conduct a detailed analysis about the challenges to apply clustering for activity recognition. Afterwards we introduce two recent subspace clustering methods that can address these challenges. Lastly, based on the analysis of unsupervised activity recognition, we will propose an activity recognition framework that incorporates clustering in conventional classification. We will show that the two directions are complementary to each other, and developing a hybrid approach will greatly benefit activity context awareness. Unsupervised activity learning |
105f3fd2054cb63223d9ffbda7b6bd5915c6be6b | Learning visual models of object categories notoriously requires thousands of training examples; this is due to the diversity and richness of object appearance which requires models containing hundreds of parameters. We present a method for learning object categories from just a few images ( ). It is based on incorporating “generic” knowledge which may be obtained from previously learnt models of unrelated categories. We operate in a variational Bayesian framework: object categories are represented by probabilistic models, and “prior” knowledge is represented as a probability density function on the parameters of these models. The “posterior” model for an object category is obtained by updating the prior in the light of one or more observations. Our ideas are demonstrated on four diverse categories (human faces, airplanes, motorcycles, spotted cats). Initially three categories are learnt from hundreds of training examples, and a “prior” is estimated from these. Then the model of the fourth category is learnt from 1 to 5 training examples, and is used for detecting new exemplars a set of test images. |
7fcd60dd4feceb576f44d138c94a04644eeb5537 | An increasing number of companies make sustainability investments, and an increasing number of investors integrate sustainability performance data in their capital allocation decisions. To date however, the prior academic literature has not distinguished between investments in material versus immaterial sustainability issues. We develop a novel dataset by hand-mapping data on sustainability investments classified as material for each industry into firm-specific performance data on a variety of sustainability investments. This allows us to present new evidence on the value implications of sustainability investments. Using calendar-time portfolio stock return regressions we find that firms with good performance on material sustainability issues significantly outperform firms with poor performance on these issues, suggesting that investments in sustainability issues are shareholder-value enhancing. Further, firms with good performance on sustainability issues not classified as material do not underperform firms with poor performance on these same issues, suggesting investments in sustainability issues are at a minimum not value-destroying. Finally, firms with good performance on material issues and concurrently poor performance on immaterial issues perform the best. These results speak to the efficiency of firms’ sustainability investments, and also have implications for asset managers who have committed to the integration of sustainability factors in their capital allocation decisions. Mozaffar Khan is a visiting Associate Professor at Harvard Business School. George Serafeim is the Jakurski Family Associate Professor of Business Administration at Harvard Business School. Aaron Yoon is a doctoral student at Harvard Business School. We are grateful for comments from seminar participants at National University of Singapore and Harvard Business School. We are grateful for financial support from the Division of Faculty Research and Development at Harvard Business School. George Serafeim has served on the Standards Council of SASB. Contact emails: [email protected]; [email protected]; [email protected]. |
65e1ada2360b42368a5f9f5e40ff436051c6fa84 | We present pomegranate, an open source machine learning package for probabilistic modeling in Python. Probabilistic modeling encompasses a wide range of methods that explicitly describe uncertainty using probability distributions. Three widely used probabilistic models implemented in pomegranate are general mixture models, hidden Markov models, and Bayesian networks. A primary focus of pomegranate is to abstract away the complexities of training models from their definition. This allows users to focus on specifying the correct model for their application instead of being limited by their understanding of the underlying algorithms. An aspect of this focus involves the collection of additive sufficient statistics from data sets as a strategy for training models. This approach trivially enables many useful learning strategies, such as out-of-core learning, minibatch learning, and semi-supervised learning, without requiring the user to consider how to partition data or modify the algorithms to handle these tasks themselves. pomegranate is written in Cython to speed up calculations and releases the global interpreter lock to allow for built-in multithreaded parallelism, making it competitive with—or outperform—other implementations of similar algorithms. This paper presents an overview of the design choices in pomegranate, and how they have enabled complex features to be supported by simple code. The code is available at https://github.com/jmschrei/pomegranate |
45fc88bbf2d007c8c8b9021d303c20611e03a715 | 0360-1315/$ see front matter 2008 Elsevier Ltd. A doi:10.1016/j.compedu.2008.06.004 * Tel.: +3 |
5ceaabef5e2ad55a781607433cc47573db786684 | The idea that a new generation of students is entering the education system has excited recent attention among educators and education commentators. Termed ‘digital natives’ or the ‘Net generation’, these young people are said to have been immersed in technology all their lives, imbuing them with sophisticated technical skills and learning preferences for which traditional education is unprepared. Grand claims are being made about the nature of this generational change and about the urgent necessity for educational reform in response. A sense of impending crisis pervades this debate. However, the actual situation is far from clear. In this paper, the authors draw on the fields of education and sociology to analyse the digital natives debate. The paper presents and questions the main claims made about digital natives and analyses the nature of the debate itself. We argue that rather than being empirically and theoretically informed, the debate can be likened to an academic form of a ‘moral panic’. We propose that a more measured and disinterested approach is now required to investigate ‘digital natives’ and their implications for education. The one thing that does not change is that at any and every time it appears that there have been ‘great changes’. Marcel Proust, Within a Budding Grove |
a708c162edd794cab982cf1049f2b81a05cf5757 | |
d36018859bbdac8797064a2f4c58e4661bbd4f02 | A Ka-band planar three-way power divider which uses the coupled line instead of the transmission line is proposed to reduce chip size. The proposed planar topology, different from the conventional Wilkinson power divider, is analyzed and can provide not only compact but also dc block characteristics, which are very suitable for monolithic microwave integrated circuit applications. The divider implemented by a pHEMT process shows an insertion loss less than 5.1 dB and an output isolation better than 17 dB. A return loss less than 18 dB and a phase difference of 4.2deg at 30 GHz can be achieved. Finally, good agreements between the simulation and experimental results are shown. |
5f9d8263ab657a9985f8f5f252a3e2da49b6e15b | This paper presents a stretchable Planar Inverted-F Antenna (PIFA) based on a new metallization technique of depositing nanoscale thin (50/100nm) Au film on elastomer Polydimethylsiloxane (PDMS). The thin metal films can be reversibly stretched up to 20% without losing electrical conduction. The PIFA antenna made of new materials can work under a 10% strain and return to its original state after removal of an applied stress. Simulation and measurement results for the antenna performance are given. |
de1a87431309ff9e9d3bce7fb0b2f3b1a62b45e3 | There is a growing interest in the integration of BIM and GIS. However, most of the research is focused on importing BIM data in GIS applications and vice versa. Real integration of BIM and GIS is using the strong parts of the GIS technology in BIM, and of course the strong parts from BIM technology in GIS. In this paper a mix of strong parts from both worlds is integrated in a single project. The paper describes the development of a CityGML extension called GeoBIM to get semantic IFC data into a GIS context. The conversion of IFC to CityGML (including the GeoBIM extension) is implemented in the open source Building Information Modelserver. This contribution was selected in a double blind review process to be published within the Lecture Notes in Geoinformation and Cartography series (Springer-Verlag, Heidelberg). Advances in 3D Geo-Information Sciences Kolbe, Thomas H.; König, Gerhard; Nagel, Claus (Eds.) 2011, X ISBN 978-3-642-12669-7, Hardcover Date of Publication: January 5, 2011 Series Editors: Cartwright, W., Gartner, G., Meng, L., Peterson, M.P. ISSN: 1863-2246 International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXVIII-4/W15 5th International 3D GeoInfo Conference, November 3-4, 2010, Berlin, Germany 193 |
4bba88ac970d88c7a4a6aa174d765649a3fcfb00 | Infusing hierarchies with elements of market control has become a much-used way of simultaneously increasing entrepreneurialism and motivation in firms. However, this paper argues that such “internal hybrids,” particularly in their radical forms, are inherently hard to successfully design and implement, because of fundamental credibility problems related to managerial promises to not intervene in delegated decision-making an incentive problem that is often referred to as the “problem of selective intervention.” This theoretical theme is developed and illustrated, using the case of the world-leading Danish hearing aids producer, Oticon. In the beginning of the 1990s, Oticon became famous for its radical internal hybrid, the ”spaghetti organization.” Recent work has interpreted the spaghetti organization as a radical attempt to foster dynamic capabilities by imposing loose coupling on the organization, neglecting, however, that about a decade later, the spaghetti organization has given way to a more traditional matrix organization. This paper presents an organizational economics interpretation of organizational changes in Oticon, and argues that a strong liability of the spaghetti organization was the above incentive problem. Motivation in Oticon was strongly harmed by selective intervention on the part of top-management Changing the organizational structure was one means of repairing these motivational problems. Refutable implications are developed, both for the understanding of efficient design of internal hybrids, and for the more general issue of the distinction between firms and markets, as well as the choice between internal and external hybrids. |
8687feb71453c7b6fb7b73802d0b77800b7285b9 | |
cfdb77b2cb9f8ca9616fa34a84d23685c0a4c45e | Reconfigurable antennas change polarization, operating frequency, or far-field pattern in order to cope with changing system parameters. This paper reviews some of the past and current technology applicable to reconfigurable antennas, with several examples of implementations. Mechanically movable parts and arrays are discussed, as well as more-recent semiconductor-component and tunable-material technologies applicable to reconfigurable antennas. |
aef5223dd82909a78e36895e1fe261d29704eb0e | A high-gain low-profile Fabry-Perot (FP) leaky-wave antenna (LWA) presenting one-dimensional high beam steering properties is proposed in this letter. The structure consists of a ground plane and a varying inductive partially reflective surface (PRS). A microstrip patch antenna is embedded into the cavity to act as the primary feed. As design examples, antennas are designed to operate at 9.5 GHz. Subwavelength FP cavities with fixed overall thickness of λ0 /6 (where λ0 is the free-space operating wavelength) are fabricated and measured. The impact of varying the PRS inductance is analyzed. It is shown that a high beam steering angle from broadside toward endfire direction close to 60 can be obtained when judiciously designing the inductive grid of the PRS. |
99d03dc3437b5c4ed1b3e54d65a030330e32c6cb | |
fef3feb39ed95a4376019af8dbfe604643fb7f25 | Stealth address prevents public association of a blockchain transaction’s output with a recipient’s wallet address and hides the actual destination address of a transaction. While stealth address provides an effective privacy-enhancing technology for a cryptocurrency network, it requires blockchain nodes to actively monitor all the transactions and compute the purported destination addresses, which restricts its application for resource-constrained environments like Internet of Things (IoT). In this paper, we propose DKSAP-IoT, a faster dual-key stealth address protocol for blockchain-based IoT systems. DKSAP-IoT utilizes a technique similar to the TLS session resumption to improve the performance and reduce the transaction size at the same time between two communication peers. Our theoretical analysis as well as the extensive experiments on an embedded computing platform demonstrate that DKSAPIoT is able to reduce the computational overhead by at least 50% when compared to the state-of-the-art scheme, thereby paving the way for its application to blockchain-based IoT systems. |
0b4b6932d5df74b366d9235b40334bc40d719c72 | In this paper, we present a simple and efficient method for training deep neural networks in a semi-supervised setting where only a small portion of training data is labeled. We introduce self-ensembling, where we form a consensus prediction of the unknown labels using the outputs of the network-in-training on different epochs, and most importantly, under different regularization and input augmentation conditions. This ensemble prediction can be expected to be a better predictor for the unknown labels than the output of the network at the most recent training epoch, and can thus be used as a target for training. Using our method, we set new records for two standard semi-supervised learning benchmarks, reducing the (non-augmented) classification error rate from 18.44% to 7.05% in SVHN with 500 labels and from 18.63% to 16.55% in CIFAR-10 with 4000 labels, and further to 5.12% and 12.16% by enabling the standard augmentations. We additionally obtain a clear improvement in CIFAR-100 classification accuracy by using random images from the Tiny Images dataset as unlabeled extra inputs during training. Finally, we demonstrate good tolerance to incorrect labels. |
Subsets and Splits