Papers.

Fast Newton active appearance models

Active Appearance Models (AAMs) are statistical models of shape and appearance widely used in computer vision to detect landmarks on objects like faces. Fitting an AAM to a new image can be formulated as a non-linear least-squares problem which is typically solved using iterative methods. Owing to its efficiency, Gauss-Newton optimization has been the standard choice over more sophisticated approaches like Newton. In this paper, we show that the AAM problem has structure which can be used to solve efficiently the original Newton problem without any approximations. We then make connections to the original Gauss-Newton algorithm and study experimentally the effect of the additional terms introduced by the Newton formulation on both fitting accuracy and convergence. Based on our derivations, we also propose a combined Newton and Gauss-Newton method which achieves promising fitting and convergence performance. Our findings are validated on two challenging in-the-wild data sets.

Pragma-based floating-to-fixed point conversion for the emulation of analog behavioral models

Highly accurate phonetic segmentation using boundary correction models and system fusion

Mandarin tone classification without pitch tracking

Automatic phonetic segmentation in Mandarin Chinese: Boundary models, glottal features and tone

Hardware system synthesis from Domain-Specific Languages

Field Programmable Gate Arrays (FPGAs) are very versatile devices, but their complicated programming model has stymied their widespread usage. While modern High-Level Synthesis (HLS) tools provide better programming models, the interface they offer is still too low-level. In order to produce good quality hardware designs with these tools, the users are forced to manually perform optimizations that demand detailed knowledge of both the application and the implementation platform. Additionally, many HLS tools only generate isolated hardware modules that the user still needs to integrate into a system design before generating the FPGA bitstream. These problems make HLS tools difficult to use for application developers who have little hardware design knowledge. To address these problems, we propose an automated methodology to generate FPGA bitstreams from high-level programs written in Domain-Specific Languages (DSLs). We leverage the domain-knowledge conveyed by the DSL and its domain-specific semantics to extract application parallelism, perform optimizations and also identify a suitable system-architecture for the implementation, thereby, relieving the user from most of the hardware-level details. We demonstrate the high productivity and high design quality this approach offers by automatically generating hardware systems from applications written in OptiML, a machine-learning DSL. To evaluate our methodology, we use four OptiML applications and show that we can easily generate different solutions which achieve different trade-offs between performance and area. More importantly, the results reveal that our generated hardware achieves much better performance compared to the one obtained from using the HLS tool without platform-specific optimizations.

Large-scale Neural Modeling in MapReduce and Giraph

Learning Everything about Anything: Webly-Supervised Visual Concept Learning

Identifying and shifting social media network patterns with NodeXL

CTS 2014 tutorials

LTE Security potential vulnerability and algorithm enhancements

Revolutionary entities: Turning data into knowledge to drive personalized exploration of The irish rising of 1916

Investigation of projection-based model-reduction techniques for solid-phase diffusion in Li-ion batteries

Mechanism for Thermal Relic Dark Matter of Strongly Interacting Massive Particles

A cloud-compatible bioinformatics pipeline for ultrarapid pathogen identification from next-generation sequencing of clinical samples

Unbiased next-generation sequencing (NGS) approaches enable comprehensive pathogen detection in the clinical microbiology laboratory and have numerous applications for public health surveillance, outbreak investigation, and the diagnosis of infectious diseases. However, practical deployment of the technology is hindered by the bioinformatics challenge of analyzing results accurately and in a clinically relevant timeframe. Here we describe SURPI (“sequence-based ultrarapid pathogen identification”), a computational pipeline for pathogen identification from complex metagenomic NGS data generated from clinical samples, and demonstrate use of the pipeline in the analysis of 237 clinical samples comprising more than 1.1 billion sequences. Deployable on both cloud-based and standalone servers, SURPI leverages two state-of-the-art aligners for accelerated analyses, SNAP and RAPSearch, which are as accurate as existing bioinformatics tools but orders of magnitude faster in performance. In fast mode, SURPI detects viruses and bacteria by scanning data sets of 7–500 million reads in 11 min to 5 h, while in comprehensive mode, all known microorganisms are identified, followed by de novo assembly and protein homology searches for divergent viruses in 50 min to 16 h. SURPI has also directly contributed to real-time microbial diagnosis in acutely ill patients, underscoring its potential key role in the development of unbiased NGS-based clinical assays in infectious diseases that demand rapid turnaround times.

Diverse Structural Evolution at z > 1 in Cosmologically Simulated Galaxies

From mock Hubble Space Telescope images, we quantify non-parametric statistics of galaxy morphology, thereby predicting the emergence of relationships among stellar mass, star formation, and observed rest-frame optical structure at 1 < z < 3. We measure automated diagnostics of galaxy morphology in cosmological simulations of the formation of 22 central galaxies with 9.3 < log10 M_*/M_sun < 10.7. These high-spatial-resolution zoom-in calculations enable accurate modeling of the rest-frame UV and optical morphology. Even with small numbers of galaxies, we find that structural evolution is neither universal nor monotonic: galaxy interactions can trigger either bulge or disc formation, and optically bulge-dominated galaxies at this mass may not remain so forever. Simulated galaxies with M_* > 10^10 M_sun contain relatively more disc-dominated light profiles than those with lower mass, reflecting significant disc brightening in some haloes at 1 < z < 2. By this epoch, simulated galaxies with specific star formation rates below 10^-9.7 yr^-1 are more likely than normal star-formers to have a broader mix of structural types, especially at M_* > 10^10 M_sun. We analyze a cosmological major merger at z ~ 1.5 and find that the newly proposed MID morphology diagnostics trace later merger stages while G-M20 trace earlier ones. MID is sensitive also to clumpy star-forming discs. The observability time of typical MID-enhanced events in our simulation sample is less than 100 Myr. A larger sample of cosmological assembly histories may be required to calibrate such diagnostics in the face of their sensitivity to viewing angle, segmentation algorithm, and various phenomena such as clumpy star formation and minor mergers.

Star Formation and Clumps in Cosmological Galaxy Simulations with Radiation Pressure Feedback

Cosmological simulations of galaxies have typically produced too many stars at early times. We study the global and morphological effects of radiation pressure (RP) in eight pairs of high-resolution cosmological galaxy formation simulations. We find that the additional feedback suppresses star formation globally by a factor of ~2. Despite this reduction, the simulations still overproduce stars by a factor of ~2 with respect to the predictions provided by abundance matching methods for halos more massive than 5E11 Msun/h (Behroozi, Wechsler & Conroy 2013). We also study the morphological impact of radiation pressure on our simulations. In simulations with RP the average number of low mass clumps falls dramatically. Only clumps with stellar masses Mclump/Mdisk <= 5% are impacted by the inclusion of RP, and RP and no-RP clump counts above this range are comparable. The inclusion of RP depresses the contrast ratios of clumps by factors of a few for clump masses less than 5% of the disk masses. For more massive clumps, the differences between and RP and no-RP simulations diminish. We note however, that the simulations analyzed have disk stellar masses below about 2E10 Msun/h. By creating mock Hubble Space Telescope observations we find that the number of clumps is slightly reduced in simulations with RP. However, since massive clumps survive the inclusion of RP and are found in our mock observations, we do not find a disagreement between simulations of our clumpy galaxies and observations of clumpy galaxies. We demonstrate that clumps found in any single gas, stellar, or mock observation image are not necessarily clumps found in another map, and that there are few clumps common to multiple maps.

Simulating multiple merger pathways to the central kinematics of early-type galaxies

Two-dimensional integral field surveys such as ATLAS^3D are producing rich observational data sets yielding insights into galaxy formation. These new kinematic observations have highlighted the need to understand the evolutionary mechanisms leading to a spectrum of fast-rotators and slow-rotators in early-type galaxies. We address the formation of slow and fast rotators through a series of controlled, comprehensive hydrodynamical simulations sampling idealized galaxy merger scenarios constructed from model spiral galaxies. Idealized and controlled simulations of this sort complement the more 'realistic' cosmological simulations by isolating and analyzing the effects of specific parameters, as we do in this paper. We recreate minor and major binary mergers, binary merger trees with multiple progenitors, and multiple sequential mergers. Within each of these categories of formation history, we correlate progenitor gas fraction, mass ratio, orbital pericenter, orbital ellipticity, and spin with remnant kinematic properties. We create kinematic profiles of these 95 simulations comparable to ATLAS^3D data. By constructing remnant profiles of the projected specific angular momentum (lambda_R = <R|V|> / <sqrt(V^2+sigma^2)>, triaxiality, and measuring the incidences of kinematic twists and kinematically decoupled cores, we distinguish between varying formation scenarios. We find that binary mergers nearly always form fast rotators. Slow rotators can be formed from zero initial angular momentum configurations and gas-poor mergers, but are not as round as the ATLAS^3D galaxies. Remnants of binary merger trees are triaxial slow rotators. Sequential mergers form round slow rotators that most resemble the ATLAS^3D rotators.

On the relationship between Gaussian stochastic blockmodels and label propagation algorithms

The problem of community detection has received great attention in recent years. Many methods have been proposed to discover communities in networks. In this paper, we propose a Gaussian stochastic blockmodel that uses Gaussian distributions to fit weight of edges in networks for non-overlapping community detection. The maximum likelihood estimation of this model has the same objective function as general label propagation with node preference. The node preference of a specific vertex turns out to be a value proportional to the intra-community eigenvector centrality (the corresponding entry in principal eigenvector of the adjacency matrix of the subgraph inside that vertex's community) under maximum likelihood estimation. Additionally, the maximum likelihood estimation of a constrained version of our model is highly related to another extension of the label propagation algorithm, namely, the label propagation algorithm under constraint. Experiments show that the proposed Gaussian stochastic blockmodel performs well on various benchmark networks.

Twitter publics: how online political communities signaled electoral outcomes in the 2010 US house election

Exome sequencing of pleuropulmonary blastoma reveals frequent biallelic loss of TP53 and two hits in DICER1 resulting in retention of 5p-derived miRNA hairpin loop sequences

RNA helicase DDX21 coordinates transcription and ribosomal RNA processing

Automatic personality assessment through social media language.

Language use is a psychologically rich, stable individual difference with well-established correlations to personality. We describe a method for assessing personality using an open-vocabulary analysis of language from social media. We compiled the written language from 66,732 Facebook users and their questionnaire-based self-reported Big Five personality traits, and then we built a predictive model of personality based on their language. We used this model to predict the 5 personality factors in a separate sample of 4,824 Facebook users, examining (a) convergence with self-reports of personality at the domain- and facet-level; (b) discriminant validity between predictions of distinct traits; (c) agreement with informant reports of personality; (d) patterns of correlations with external criteria (e.g., number of friends, political attitudes, impulsiveness); and (e) test-retest reliability over 6-month intervals. Results indicated that language-based assessments can constitute valid personality measures: they agreed with self-reports and informant reports of personality, added incremental validity over informant reports, adequately discriminated between traits, exhibited patterns of correlations with external criteria similar to those found with self-reported personality, and were stable over 6-month intervals. Analysis of predictive language can provide rich portraits of the mental life associated with traits. This approach can complement and extend traditional methods, providing researchers with an additional measure that can quickly and cheaply assess large groups of participants with minimal burden.

Introduction to the art of programming using Scala, by Mark C. Lewis, Chapman and Hall/CRC Press, 2012, £ 46.99 (paperback) ISBN-10:1439896666

F0 declination in English and Mandarin Broadcast News Speech

Automatic multi-partite graph generation from arbitrary data

Approximate nearest neighbor algorithm based on navigable small world graphs

Crude oil: Commodity or financial asset?

Intra-layer network coding for lossy communication of progressive codes

ImageNet Large Scale Visual Recognition Challenge

The ImageNet Large Scale Visual Recognition Challenge is a benchmark in object category classification and detection on hundreds of object categories and millions of images. The challenge has been run annually from 2010 to present, attracting participation from more than fifty institutions. This paper describes the creation of this benchmark dataset and the advances in object recognition that have been possible as a result. We discuss the challenges of collecting large-scale ground truth annotation, highlight key breakthroughs in categorical object recognition, provide a detailed analysis of the current state of the field of large-scale image classification and object detection, and compare the state-of-the-art computer vision accuracy with human accuracy. We conclude with lessons learned in the five years of the challenge, and propose future directions and improvements.

Evader interdiction: algorithms, complexity and collateral damage

Distributed constraint optimization for teams of mobile sensing agents

The Stratosphere platform for big data analytics

Boosting Stop Searches with a 100 TeV Proton Collider

Jet Substructure Templates: Data-driven QCD Backgrounds for Fat Jet Searches

SUSY Simplified Models at 14, 33, and 100 TeV Proton Colliders

Spores: A Type-Based Foundation for Closures in the Age of Concurrency and Distribution

Scaling HDFS with a Strongly Consistent Relational Model for Metadata

Towards a Peer-to-Peer Bandwidth Marketplace

Who Are My Audiences? A Study of the Evolution of Target Audiences in Microblogs

Diachronic Deviation Features in Continuous Space Word Representations

Advancing Big Data Benchmarks

Key Management for Onion Routing in a True Peer to Peer Setting

Prescient Profiling – AI Driven Volunteer Selection within a Volunteer Notification System

Towards Personalized Multilingual Information Access - Exploring the Browsing and Search Behavior of Multilingual Users

Crowdsourcing and Human Computation: Introduction

Social Network Analysis in HCI

Hidden Aspects of a Cloud Computing Contract

Assessing Cloud Computing for Business Use

Cloud Computing Security