Showing papers by "Virginia Tech published in 2016"

Hierarchical Question-Image Co-Attention for Visual Question Answering

Abstract: A number of recent works have proposed attention models for Visual Question Answering (VQA) that generate spatial maps highlighting image regions relevant to answering the question. In this paper, we argue that in addition to modeling "where to look" or visual attention, it is equally important to model "what words to listen to" or question attention. We present a novel co-attention model for VQA that jointly reasons about image and question attention. In addition, our model reasons about the question (and consequently the image via the co-attention mechanism) in a hierarchical fashion via a novel 1-dimensional convolution neural networks (CNN). Our model improves the state-of-the-art on the VQA dataset from 60.3% to 60.5%, and from 61.6% to 63.3% on the COCO-QA dataset. By using ResNet, the performance is further improved to 62.1% for VQA and 65.4% for COCO-QA.

Unmanned Aerial Vehicle With Underlaid Device-to-Device Communications: Performance and Tradeoffs

Abstract: In this paper, the deployment of an unmanned aerial vehicle (UAV) as a flying base station used to provide the fly wireless communications to a given geographical area is analyzed. In particular, the coexistence between the UAV, that is transmitting data in the downlink, and an underlaid device-to-device (D2D) communication network is considered. For this model, a tractable analytical framework for the coverage and rate analysis is derived. Two scenarios are considered: a static UAV and a mobile UAV. In the first scenario, the average coverage probability and the system sum-rate for the users in the area are derived as a function of the UAV altitude and the number of D2D users. In the second scenario, using the disk covering problem, the minimum number of stop points that the UAV needs to visit in order to completely cover the area is computed. Furthermore, considering multiple retransmissions for the UAV and D2D users, the overall outage probability of the D2D users is derived. Simulation and analytical results show that, depending on the density of D2D users, the optimal values for the UAV altitude, which lead to the maximum system sum-rate and coverage probability, exist. Moreover, our results also show that, by enabling the UAV to intelligently move over the target area, the total required transmit power of UAV while covering the entire area, can be minimized. Finally, in order to provide full coverage for the area of interest, the tradeoff between the coverage and delay, in terms of the number of stop points, is discussed.

Efficient Deployment of Multiple Unmanned Aerial Vehicles for Optimal Wireless Coverage

Abstract: In this letter, the efficient deployment of multiple unmanned aerial vehicles (UAVs) acting as wireless base stations that provide coverage for ground users is analyzed. First, the downlink coverage probability for UAVs as a function of the altitude and the antenna gain is derived. Next, using circle packing theory, the 3-D locations of the UAVs is determined in a way that the total coverage area is maximized while maximizing the coverage lifetime of the UAVs. Our results show that, in order to mitigate interference, the altitude of the UAVs must be properly adjusted based on the beamwidth of the directional antenna as well as coverage requirements. Furthermore, the minimum number of UAVs needed to guarantee a target coverage probability for a given geographical area is determined. Numerical results evaluate various tradeoffs.

Knowing When to Look: Adaptive Attention via A Visual Sentinel for Image Captioning

Abstract: Attention-based neural encoder-decoder frameworks have been widely adopted for image captioning. Most methods force visual attention to be active for every generated word. However, the decoder likely requires little to no visual information from the image to predict non-visual words such as "the" and "of". Other words that may seem visual can often be predicted reliably just from the language model e.g., "sign" after "behind a red stop" or "phone" following "talking on a cell". In this paper, we propose a novel adaptive attention model with a visual sentinel. At each time step, our model decides whether to attend to the image (and if so, to which regions) or to the visual sentinel. The model decides whether to attend to the image and where, in order to extract meaningful information for sequential word generation. We test our method on the COCO image captioning 2015 challenge dataset and Flickr30K. Our approach sets the new state-of-the-art by a significant margin.

Analysis of D-Q Small-Signal Impedance of Grid-Tied Inverters

Abstract: This paper analyzes the small-signal impedance of three-phase grid-tied inverters with feedback control and phase-locked loop (PLL) in the synchronous reference ( d-q ) frame. The result unveils an interesting and important feature of three-phase grid-tied inverters – namely, that its q–q channel impedance behaves as a negative incremental resistor. Moreover, this paper shows that this behavior is a consequence of grid synchronization, where the bandwidth of the PLL determines the frequency range of the resistor behavior, and the power rating of the inverter determines the magnitude of the resistor. Advanced PLL, current, and power control strategies do not change this feature. An example shows that under weak grid conditions, a change of the PLL bandwidth could lead the inverter system to unstable conditions as a result of this behavior. Harmonic resonance and instability issues can be analyzed using the proposed impedance model. Simulation and experimental measurements verify the analysis.

Convolutional Radio Modulation Recognition Networks

Abstract: We study the adaptation of convolutional neural networks to the complex-valued temporal radio signal domain. We compare the efficacy of radio modulation classification using naively learned features against using expert feature based methods which are widely used today and e show significant performance improvements. We show that blind temporal learning on large and densely encoded time series using deep convolutional neural networks is viable and a strong candidate approach for this task especially at low signal to noise ratio.

Association Between Palliative Care and Patient and Caregiver Outcomes: A Systematic Review and Meta-analysis.

Abstract: Importance The use of palliative care programs and the number of trials assessing their effectiveness have increased. Objective To determine the association of palliative care with quality of life (QOL), symptom burden, survival, and other outcomes for people with life-limiting illness and for their caregivers. Data Sources MEDLINE, EMBASE, CINAHL, and Cochrane CENTRAL to July 2016. Study Selection Randomized clinical trials of palliative care interventions in adults with life-limiting illness. Data Extraction and Synthesis Two reviewers independently extracted data. Narrative synthesis was conducted for all trials. Quality of life, symptom burden, and survival were analyzed using random-effects meta-analysis, with estimates of QOL translated to units of the Functional Assessment of Chronic Illness Therapy–palliative care scale (FACIT-Pal) instrument (range, 0-184 [worst-best]; minimal clinically important difference [MCID], 9 points); and symptom burden translated to the Edmonton Symptom Assessment Scale (ESAS) (range, 0-90 [best-worst]; MCID, 5.7 points). Main Outcomes and Measures Quality of life, symptom burden, survival, mood, advance care planning, site of death, health care satisfaction, resource utilization, and health care expenditures. Results Forty-three RCTs provided data on 12 731 patients (mean age, 67 years) and 2479 caregivers. Thirty-five trials used usual care as the control, and 14 took place in the ambulatory setting. In the meta-analysis, palliative care was associated with statistically and clinically significant improvements in patient QOL at the 1- to 3-month follow-up (standardized mean difference, 0.46; 95% CI, 0.08 to 0.83; FACIT-Pal mean difference, 11.36] and symptom burden at the 1- to 3-month follow-up (standardized mean difference, −0.66; 95% CI, −1.25 to −0.07; ESAS mean difference, −10.30). When analyses were limited to trials at low risk of bias (n = 5), the association between palliative care and QOL was attenuated but remained statistically significant (standardized mean difference, 0.20; 95% CI, 0.06 to 0.34; FACIT-Pal mean difference, 4.94), whereas the association with symptom burden was not statistically significant (standardized mean difference, −0.21; 95% CI, −0.42 to 0.00; ESAS mean difference, −3.28). There was no association between palliative care and survival (hazard ratio, 0.90; 95% CI, 0.69 to 1.17). Palliative care was associated consistently with improvements in advance care planning, patient and caregiver satisfaction, and lower health care utilization. Evidence of associations with other outcomes was mixed. Conclusions and Relevance In this meta-analysis, palliative care interventions were associated with improvements in patient QOL and symptom burden. Findings for caregiver outcomes were inconsistent. However, many associations were no longer significant when limited to trials at low risk of bias, and there was no significant association between palliative care and survival.

Driver crash risk factors and prevalence evaluation using naturalistic driving data.

Abstract: The accurate evaluation of crash causal factors can provide fundamental information for effective transportation policy, vehicle design, and driver education. Naturalistic driving (ND) data collected with multiple onboard video cameras and sensors provide a unique opportunity to evaluate risk factors during the seconds leading up to a crash. This paper uses a National Academy of Sciences-sponsored ND dataset comprising 905 injurious and property damage crash events, the magnitude of which allows the first direct analysis (to our knowledge) of causal factors using crashes only. The results show that crash causation has shifted dramatically in recent years, with driver-related factors (i.e., error, impairment, fatigue, and distraction) present in almost 90% of crashes. The results also definitively show that distraction is detrimental to driver safety, with handheld electronic devices having high use rates and risk.

Sparse maps--A systematic infrastructure for reduced-scaling electronic structure methods. II. Linear scaling domain based pair natural orbital coupled cluster theory.

Abstract: Domain based local pair natural orbital coupled cluster theory with single-, double-, and perturbative triple excitations (DLPNO-CCSD(T)) is a highly efficient local correlation method. It is known to be accurate and robust and can be used in a black box fashion in order to obtain coupled cluster quality total energies for large molecules with several hundred atoms. While previous implementations showed near linear scaling up to a few hundred atoms, several nonlinear scaling steps limited the applicability of the method for very large systems. In this work, these limitations are overcome and a linear scaling DLPNO-CCSD(T) method for closed shell systems is reported. The new implementation is based on the concept of sparse maps that was introduced in Part I of this series [P. Pinski, C. Riplinger, E. F. Valeev, and F. Neese, J. Chem. Phys. 143, 034108 (2015)]. Using the sparse map infrastructure, all essential computational steps (integral transformation and storage, initial guess, pair natural orbital construction, amplitude iterations, triples correction) are achieved in a linear scaling fashion. In addition, a number of additional algorithmic improvements are reported that lead to significant speedups of the method. The new, linear-scaling DLPNO-CCSD(T) implementation typically is 7 times faster than the previous implementation and consumes 4 times less disk space for large three-dimensional systems. For linear systems, the performance gains and memory savings are substantially larger. Calculations with more than 20 000 basis functions and 1000 atoms are reported in this work. In all cases, the time required for the coupled cluster step is comparable to or lower than for the preceding Hartree-Fock calculation, even if this is carried out with the efficient resolution-of-the-identity and chain-of-spheres approximations. The new implementation even reduces the error in absolute correlation energies by about a factor of two, compared to the already accurate previous implementation.

Joint Unsupervised Learning of Deep Representations and Image Clusters

Abstract: In this paper, we propose a recurrent framework for joint unsupervised learning of deep representations and image clusters. In our framework, successive operations in a clustering algorithm are expressed as steps in a recurrent process, stacked on top of representations output by a Convolutional Neural Network (CNN). During training, image clusters and representations are updated jointly: image clustering is conducted in the forward pass, while representation learning in the backward pass. Our key idea behind this framework is that good representations are beneficial to image clustering and clustering results provide supervisory signals to representation learning. By integrating two processes into a single model with a unified weighted triplet loss function and optimizing it end-to-end, we can obtain not only more powerful representations, but also more precise image clusters. Extensive experiments show that our method outperforms the state of-the-art on image clustering across a variety of image datasets. Moreover, the learned representations generalize well when transferred to other tasks. The source code can be downloaded from https://github.com/ jwyang/joint-unsupervised-learning.

Origin of J–V Hysteresis in Perovskite Solar Cells

Abstract: High-performance perovskite solar cells (PSCs) based on organometal halide perovskite have emerged in the past five years as excellent devices for harvesting solar energy. Some remaining challenges should be resolved to continue the momentum in their development. The photocurrent density–voltage (J−V) responses of the PSCs demonstrate anomalous dependence on the voltage scan direction/rate/range, voltage conditioning history, and device configuration. The hysteretic J–V behavior presents a challenge for determining the accurate power conversion efficiency of the PSCs. Here, we review the recent progress on the investigation of the origin(s) of J–V hysteresis behavior in PSCs. We discuss the impact of slow transient capacitive current, trapping and detrapping process, ion migrations, and ferroelectric polarization on the hysteresis behavior. The remaining issues and future research required toward the understanding of J–V hysteresis in PSCs will also be discussed.

A Corpus and Cloze Evaluation for Deeper Understanding of Commonsense Stories

Abstract: Representation and learning of commonsense knowledge is one of the foundational problems in the quest to enable deep language understanding. This issue is particularly challenging for understanding casual and correlational relationships between events. While this topic has received a lot of interest in the NLP community, research has been hindered by the lack of a proper evaluation framework. This paper attempts to address this problem with a new framework for evaluating story understanding and script learning: the `Story Cloze Test’. This test requires a system to choose the correct ending to a four-sentence story. We created a new corpus of 50k five-sentence commonsense stories, ROCStories, to enable this evaluation. This corpus is unique in two ways: (1) it captures a rich set of causal and temporal commonsense relations between daily events, and (2) it is a high quality collection of everyday life stories that can also be used for story generation. Experimental evaluation shows that a host of baselines and state-of-the-art models based on shallow language understanding struggle to achieve a high score on the Story Cloze Test. We discuss these implications for script and story learning, and offer suggestions for deeper language understanding.

Multiscale metallic metamaterials

Abstract: Materials with three-dimensional micro- and nanoarchitectures exhibit many beneficial mechanical, energy conversion and optical properties. However, these three-dimensional microarchitectures are significantly limited by their scalability. Efforts have only been successful only in demonstrating overall structure sizes of hundreds of micrometres, or contain size-scale gaps of several orders of magnitude. This results in degraded mechanical properties at the macroscale. Here we demonstrate hierarchical metamaterials with disparate three-dimensional features spanning seven orders of magnitude, from nanometres to centimetres. At the macroscale they achieve high tensile elasticity (>20%) not found in their brittle-like metallic constituents, and a near-constant specific strength. Creation of these materials is enabled by a high-resolution, large-area additive manufacturing technique with scalability not achievable by two-photon polymerization or traditional stereolithography. With overall part sizes approaching tens of centimetres, these unique nanostructured metamaterials might find use in a broad array of applications.

Unified Theory of Acceptance and Use of Technology: A Synthesis and the Road Ahead

Abstract: The unified theory of acceptance and use of technology (UTAUT) is a little over a decade old and has been used extensively in information systems (IS) and other fields, as the large number of citations to the original paper that introduced the theory evidences. In this paper, we review and synthesize the IS literature on UTAUT from September 2003 until December 2014, perform a theoretical analysis of UTAUT and its extensions, and chart an agenda for research going forward. Based on Weber’s (2012) framework of theory evaluation, we examined UTAUT and its extensions along two sets of quality dimensions; namely, the parts of a theory and the theory as a whole. While our review identifies many merits to UTAUT, we also found that the progress related to this theory has hampered further theoretical development in research into technology acceptance and use. To chart an agenda for research that will enable significant future work, we analyze the theoretical contributions of UTAUT using Whetten’s (2009) notion of cross-context theorizing. Our analysis reveals several limitations that lead us to propose a multi-level framework that can serve as the theoretical foundation for future research. Specifically, this framework integrates the notion of research context and cross-context theorizing with the theory evaluation framework to: (1) synthesize the existing UTAUT extensions across both the dimensions and the levels of the research context and (2) highlight promising research directions. We conclude with recommendations for future UTAUT-related research using the proposed framework.

Hierarchical Question-Image Co-Attention for Visual Question Answering

Abstract: A number of recent works have proposed attention models for Visual Question Answering (VQA) that generate spatial maps highlighting image regions relevant to answering the question. In this paper, we argue that in addition to modeling "where to look" or visual attention, it is equally important to model "what words to listen to" or question attention. We present a novel co-attention model for VQA that jointly reasons about image and question attention. In addition, our model reasons about the question (and consequently the image via the co-attention mechanism) in a hierarchical fashion via a novel 1-dimensional convolution neural networks (CNN). Our model improves the state-of-the-art on the VQA dataset from 60.3% to 60.5%, and from 61.6% to 63.3% on the COCO-QA dataset. By using ResNet, the performance is further improved to 62.1% for VQA and 65.4% for COCO-QA.

Microbes as engines of ecosystem function: When does community structure enhance predictions of ecosystem processes?

Abstract: Microorganisms are vital in mediating the earth's biogeochemical cycles; yet, despite our rapidly increasing ability to explore complex environmental microbial communities, the relationship between microbial community structure and ecosystem processes remains poorly understood. Here, we address a fundamental and unanswered question in microbial ecology: 'When do we need to understand microbial community structure to accurately predict function?' We present a statistical analysis investigating the value of environmental data and microbial community structure independently and in combination for explaining rates of carbon and nitrogen cycling processes within 82 global datasets. Environmental variables were the strongest predictors of process rates but left 44% of variation unexplained on average, suggesting the potential for microbial data to increase model accuracy. Although only 29% of our datasets were significantly improved by adding information on microbial community structure, we observed improvement in models of processes mediated by narrow phylogenetic guilds via functional gene data, and conversely, improvement in models of facultative microbial processes via community diversity metrics. Our results also suggest that microbial diversity can strengthen predictions of respiration rates beyond microbial biomass parameters, as 53% of models were improved by incorporating both sets of predictors compared to 35% by microbial biomass alone. Our analysis represents the first comprehensive analysis of research examining links between microbial community structure and ecosystem function. Taken together, our results indicate that a greater understanding of microbial communities informed by ecological principles may enhance our ability to predict ecosystem process rates relative to assessments based on environmental variables and microbial physiology.

Dark Matter Search Results from the PICO-2L C3F8 Bubble Chamber.

Abstract: New results are reported from the operation of the PICO-60 dark matter detector, a bubble chamber filled with 52 kg of C_{3}F_{8} located in the SNOLAB underground laboratory. As in previous PICO bubble chambers, PICO-60 C_{3}F_{8} exhibits excellent electron recoil and alpha decay rejection, and the observed multiple-scattering neutron rate indicates a single-scatter neutron background of less than one event per month. A blind analysis of an efficiency-corrected 1167-kg day exposure at a 3.3-keV thermodynamic threshold reveals no single-scattering nuclear recoil candidates, consistent with the predicted background. These results set the most stringent direct-detection constraint to date on the weakly interacting massive particle (WIMP)-proton spin-dependent cross section at 3.4×10^{-41} cm^{2} for a 30-GeV c^{-2} WIMP, more than 1 order of magnitude improvement from previous PICO results.

Smartphone Use in Everyday Life and Travel

Abstract: The smartphone penetrates many facets of everyday life, including travel. As such, this article argues that since travel can be considered a special stage of technology use, understanding how the smartphone shapes the tourist experience cannot be separated from the way it is used in one’s everyday life. On the basis of a study of American travelers, this study uses adaptive structuration theory as a lens to identify a number of spillover effects from smartphone use in everyday life into travel. The results of this study offer several important implications for both research and practice as well as future directions for the study of mobile technology in tourism.

Genomic insights into the Ixodes scapularis tick vector of Lyme disease

Abstract: Ticks transmit more pathogens to humans and animals than any other arthropod. We describe the 2.1 Gbp nuclear genome of the tick, Ixodes scapularis (Say), which vectors pathogens that cause Lyme disease, human granulocytic anaplasmosis, babesiosis and other diseases. The large genome reflects accumulation of repetitive DNA, new lineages of retro-transposons, and gene architecture patterns resembling ancient metazoans rather than pancrustaceans. Annotation of scaffolds representing ∼57% of the genome, reveals 20,486 protein-coding genes and expansions of gene families associated with tick-host interactions. We report insights from genome analyses into parasitic processes unique to ticks, including host 'questing', prolonged feeding, cuticle synthesis, blood meal concentration, novel methods of haemoglobin digestion, haem detoxification, vitellogenesis and prolonged off-host survival. We identify proteins associated with the agent of human granulocytic anaplasmosis, an emerging disease, and the encephalitis-causing Langat virus, and a population structure correlated to life-history traits and transmission of the Lyme disease agent.