Showing papers by "University of Texas at Arlington published in 2019"

DropEdge: Towards Deep Graph Convolutional Networks on Node Classification.

Abstract: \emph{Over-fitting} and \emph{over-smoothing} are two main obstacles of developing deep Graph Convolutional Networks (GCNs) for node classification. In particular, over-fitting weakens the generalization ability on small dataset, while over-smoothing impedes model training by isolating output representations from the input features with the increase in network depth. This paper proposes DropEdge, a novel and flexible technique to alleviate both issues. At its core, DropEdge randomly removes a certain number of edges from the input graph at each training epoch, acting like a data augmenter and also a message passing reducer. Furthermore, we theoretically demonstrate that DropEdge either reduces the convergence speed of over-smoothing or relieves the information loss caused by it. More importantly, our DropEdge is a general skill that can be equipped with many other backbone models (e.g. GCN, ResGCN, GraphSAGE, and JKNet) for enhanced performance. Extensive experiments on several benchmarks verify that DropEdge consistently improves the performance on a variety of both shallow and deep GCNs. The effect of DropEdge on preventing over-smoothing is empirically visualized and validated as well. Codes are released on~\url{this https URL}.

Graph Convolutional Networks for Temporal Action Localization

Abstract: Most state-of-the-art action localization systems process each action proposal individually, without explicitly exploiting their relations during learning. However, the relations between proposals actually play an important role in action localization, since a meaningful action always consists of multiple proposals in a video. In this paper, we propose to exploit the proposal-proposal relations using GraphConvolutional Networks (GCNs). First, we construct an action proposal graph, where each proposal is represented as a node and their relations between two proposals as an edge. Here, we use two types of relations, one for capturing the context information for each proposal and the other one for characterizing the correlations between distinct actions. Then we apply the GCNs over the graph to model the relations among different proposals and learn powerful representations for the action classification and localization. Experimental results show that our approach significantly outperforms the state-of-the-art on THUMOS14(49.1% versus 42.8%). Moreover, augmentation experiments on ActivityNet also verify the efficacy of modeling action proposal relationships.

Progressive Feature Alignment for Unsupervised Domain Adaptation

Abstract: Unsupervised domain adaptation (UDA) transfers knowledge from a label-rich source domain to a fully-unlabeled target domain. To tackle this task, recent approaches resort to discriminative domain transfer in virtue of pseudo-labels to enforce the class-level distribution alignment across the source and target domains. These methods, however, are vulnerable to the error accumulation and thus incapable of preserving cross-domain category consistency, as the pseudo-labeling accuracy is not guaranteed explicitly. In this paper, we propose the Progressive Feature Alignment Network (PFAN) to align the discriminative features across domains progressively and effectively, via exploiting the intra-class variation in the target domain. To be specific, we first develop an Easy-to-Hard Transfer Strategy (EHTS) and an Adaptive Prototype Alignment (APA) step to train our model iteratively and alternatively. Moreover, upon observing that a good domain adaptation usually requires a non-saturated source classifier, we consider a simple yet efficient way to retard the convergence speed of the source classification loss by further involving a temperature variate into the soft-max function. The extensive experimental results reveal that the proposed PFAN exceeds the state-of-the-art performance on three UDA datasets.

Third generation of vortex identification methods: Omega and Liutex/Rortex based systems

Abstract: A vortex is intuitively recognized as the rotational/swirling motion of fluids, but a rigorous and universally-accepted definition is still not available. Vorticity tube/filament has been regarded equivalent to a vortex since Helmholtz proposed the concepts of vorticity tube/filament in 1858 and the vorticity-based methods can be categorized as the first generation of vortex identification methods. During the last three decades, a lot of vortex identification methods, including Q, Δ, λ2 and λci criteria, have been proposed to overcome the problems associated with the vorticity-based methods. Most of these criteria are based on the Cauchy-Stokes decomposition and/or eigenvalues of the velocity gradient tensor and can be considered as the second generation of vortex identification methods. Starting from 2014, the Vortex and Turbulence Research Team at the University of Texas at Arlington (the UTA team) focus on the development of a new generation of vortex identification methods. The first fruit of this effort, a new omega (Ω) vortex identification method, which defined a vortex as a connected region where the vorticity overtakes the deformation, was published in 2016. In 2017 and 2018, a Liutex (previously called Rortex) vector was proposed to provide a mathematical definition of the local rigid rotation part of the fluid motion, including both the local rotational axis and the rotational strength. Liutex/Rortex is a new physical quantity with scalar, vector and tensor forms exactly representing the local rigid rotation of fluids. Meanwhile, a decomposition of the vorticity to a rotational part namely Liutex/Rortex and an anti-symmetric shear part (RS decomposition) was introduced in 2018, and a universal decomposition of the velocity gradient tensor to a rotation part (R) and a non-rotation part (NR) was also given in 2018 as a counterpart of the traditional Cauchy-Stokes decomposition. Later in early 2019, a Liutex/Rortex based Omega method called Omega-Liutex, which combines the respective advantages of both Liutex/Rortex and Omega methods, was developed. And a latest objective Omega method, which is still under development, is also briefly introduced. These advances are classified as the third generation of vortex identification methods in the current paper. To elaborate the advantages of the third-generation methods, six core issues for vortex definition and identification have been raised, including: (1) the absolute strength, (2) the relative strength, (3) the rotational axis, (4) the vortex core center location, (5) the vortex core size, (6) the vortex boundary. The new third generation of vortex identification methods can provide reasonable answers to these questions, while other vortex identification methods fail to answer all questions except for the approximation of vortex boundaries. The purpose of the current paper is to summarize the main ideas and methods of the third generation of vortex identification methods rather than to conduct a comprehensive review on the historical development of vortex identification methods.

Combination of searches for invisible Higgs boson decays with the ATLAS experiment

Abstract: Dark matter particles, if sufficiently light, may be produced in decays of the Higgs boson. This Letter presents a statistical combination of searches for H→invisible decays where H is produced according to the standard model via vector boson fusion, Z(ll)H, and W/Z(had)H, all performed with the ATLAS detector using 36.1 fb^{-1} of pp collisions at a center-of-mass energy of sqrt[s]=13 TeV at the LHC. In combination with the results at sqrt[s]=7 and 8 TeV, an exclusion limit on the H→invisible branching ratio of 0.26(0.17_{-0.05}^{+0.07}) at 95% confidence level is observed (expected).

Electron and photon performance measurements with the ATLAS detector using the 2015-2017 LHC proton-proton collision data

Abstract: This paper describes the reconstruction of electrons and photons with the ATLAS detector, employed for measurements and searches exploiting the complete LHC Run 2 dataset. An improved energy clustering algorithm is introduced, and its implications for the measurement and identification of prompt electrons and photons are discussed in detail. Corrections and calibrations that affect performance, including energy calibration, identification and isolation efficiencies, and the measurement of the charge of reconstructed electron candidates are determined using up to 81 fb−1 of proton-proton collision data collected at √s=13 TeV between 2015 and 2017.

ATLAS b-jet identification performance and efficiency measurement with tt¯ events in pp collisions at √s = 13 TeV

Abstract: The algorithms used by the ATLAS Collaboration during Run 2 of the Large Hadron Collider to identify jets containing b-hadrons are presented. The performance of the algorithms is evaluated in the s ...

Study of the rare decays of B s 0 and B 0 mesons into muon pairs using data collected during 2015 and 2016 with the ATLAS detector

Abstract: A study of the decays B0 s ! + and B0 ! + has been performed using 26 : 3 fb of 13TeV LHC proton-proton collision data collected with the ATLAS detector in 2015 and 2016. Since the detector resolut ...

SMILES-BERT: Large Scale Unsupervised Pre-Training for Molecular Property Prediction

Abstract: With the rapid progress of AI in both academia and industry, Deep Learning has been widely introduced into various areas in drug discovery to accelerate its pace and cut R&D costs. Among all the problems in drug discovery, molecular property prediction has been one of the most important problems. Unlike general Deep Learning applications, the scale of labeled data is limited in molecular property prediction. To better solve this problem, Deep Learning methods have started focusing on how to utilize tremendous unlabeled data to improve the prediction performance on small-scale labeled data. In this paper, we propose a semi-supervised model named SMILES-BERT, which consists of attention mechanism based Transformer Layer. A large-scale unlabeled data has been used to pre-train the model through a Masked SMILES Recovery task. Then the pre-trained model could easily be generalized into different molecular property prediction tasks via fine-tuning. In the experiments, the proposed SMILES-BERT outperforms the state-of-the-art methods on all three datasets, showing the effectiveness of our unsupervised pre-training and great generalization capability of the pre-trained model.

Age and pattern of the southern high-latitude continental end-Permian extinction constrained by multiproxy analysis

Abstract: Past studies of the end-Permian extinction (EPE), the largest biotic crisis of the Phanerozoic, have not resolved the timing of events in southern high-latitudes. Here we use palynology coupled with high-precision CA-ID-TIMS dating of euhedral zircons from continental sequences of the Sydney Basin, Australia, to show that the collapse of the austral Permian Glossopteris flora occurred prior to 252.3 Ma (~370 kyrs before the main marine extinction). Weathering proxies indicate that floristic changes occurred during a brief climate perturbation in a regional alluvial landscape that otherwise experienced insubstantial change in fluvial style, insignificant reorganization of the depositional surface, and no abrupt aridification. Palaeoclimate modelling suggests a moderate shift to warmer summer temperatures and amplified seasonality in temperature across the EPE, and warmer and wetter conditions for all seasons into the Early Triassic. The terrestrial EPE and a succeeding peak in Ni concentration in the Sydney Basin correlate, respectively, to the onset of the primary extrusive and intrusive phases of the Siberian Traps Large Igneous Province. The continental record of the end Permian mass extinction is limited, especially from high paleolatitudes. Here, Fielding et al. report a multi-proxy Permo-Triassic record from Australia, resolving the timing of local terrestrial plant extinction and the relationship with environmental changes.

lambda-Net: Reconstruct Hyperspectral Images From a Snapshot Measurement

Abstract: We propose the λ-net, which reconstructs hyperspectral images (e.g., with 24 spectral channels) from a single shot measurement. This task is usually termed snapshot compressive-spectral imaging (SCI), which enjoys low cost, low bandwidth and high-speed sensing rate via capturing the three-dimensional (3D) signal i.e., (x, y, λ), using a 2D snapshot. Though proposed more than a decade ago, the poor quality and low-speed of reconstruction algorithms preclude wide applications of SCI. To address this challenge, in this paper, we develop a dual-stage generative model to reconstruct the desired 3D signal in SCI, dubbed λ-net. Results on both simulation and real datasets demonstrate the significant advantages of λ-net, which leads to >4dB improvement in PSNR for real-mask-in-the-loop simulation data compared to the current state-of-the-art. Furthermore, λ-net can finish the reconstruction task within sub-seconds instead of hours taken by the most recently proposed DeSCI algorithm, thus speeding up the reconstruction >1000 times.

Electron reconstruction and identification in the ATLAS experiment using the 2015 and 2016 LHC proton-proton collision data at s=13 TeV

Abstract: Algorithms used for the reconstruction and identification of electrons in the central region of the ATLAS detector at the Large Hadron Collider (LHC) are presented in this paper; these algorithms a ...

Achieving superelasticity in additively manufactured NiTi in compression without post-process heat treatment

Abstract: Shape memory alloys (SMAs), such as Nitinol (i.e., NiTi), are of great importance in biomedical and engineering applications due to their unique superelasticity and shape memory properties. In recent years, additive manufacturing (AM) processes have been used to produce complex NiTi components, which provide the ability to tailor microstructure and thus the critical properties of the alloys, such as the superelastic behavior and transformation temperatures (TTs), by selection of processing parameters. In biomedical applications, superelasticity in implants play a critical role since it gives the implants bone-like behavior. In this study, a methodology of improving superelasticity in Ni-rich NiTi components without the need for any kind of post-process heat treatments will be revealed. It will be shown that superelasticity with 5.62% strain recovery and 98% recovery ratio can be observed in Ni-rich NiTi after the sample is processed with 250 W laser power, 1250 mm/s scanning speed, and 80 µm hatch spacing without, any post-process heat treatments. This superelasticity in as-fabricated Ni-rich SLM NiTi was not previously possible in the absence of post-process heat treatments. The findings of this study promise the fast, reliable and inexpensive fabrication of complex shaped superelastic NiTi components for many envisioned applications such as patient-specific biomedical implants.

Proppy: Organizing the news based on their propagandistic content

Abstract: Propaganda is a mechanism to influence public opinion, which is inherently present in extremely biased and fake news. Here, we propose a model to automatically assess the level of propagandistic content in an article based on different representations, from writing style and readability level to the presence of certain keywords. We experiment thoroughly with different variations of such a model on a new publicly available corpus, and we show that character n-grams and other style features outperform existing alternatives to identify propaganda based on word n-grams. Unlike previous work, we make sure that the test data comes from news sources that were unseen on training, thus penalizing learning algorithms that model the news sources used at training time as opposed to solving the actual task. We integrate our supervised model in a public website, which organizes recent articles covering the same event on the basis of their propagandistic contents. This allows users to quickly explore different perspectives of the same story, and it also enables investigative journalists to dig further into how different media use stories and propaganda to pursue their agenda.

Designing Fair Ranking Schemes

Abstract: Items from a database are often ranked based on a combination of criteria. The weight given to each criterion in the combination can greatly affect the fairness of the produced ranking, for example, preferring men over women. A user may have the flexibility to choose combinations that weigh these criteria differently, within limits. In this paper, we develop a system that helps users choose criterion weights that lead to greater fairness. We consider ranking functions that compute the score of each item as a weighted sum of (numeric) attribute values, and then sort items on their score. Each ranking function can be expressed as a point in a multi-dimensional space. For a broad range of fairness criteria, including proportionality, we show how to efficiently identify regions in this space that satisfy these criteria. Using this identification method, our system is able to tell users whether their proposed ranking function satisfies the desired fairness criteria and, if it does not, to suggest the smallest modification that does. Our extensive experiments on real datasets demonstrate that our methods are able to find solutions that satisfy fairness criteria effectively (usually with only small changes to proposed weight vectors) and efficiently (in interactive time, after some initial pre-processing).

Bioresorbable optical sensor systems for monitoring of intracranial pressure and temperature

Abstract: Continuous measurements of pressure and temperature within the intracranial, intraocular, and intravascular spaces provide essential diagnostic information for the treatment of traumatic brain injury, glaucoma, and cardiovascular diseases, respectively. Optical sensors are attractive because of their inherent compatibility with magnetic resonance imaging (MRI). Existing implantable optical components use permanent, nonresorbable materials that must be surgically extracted after use. Bioresorbable alternatives, introduced here, bypass this requirement, thereby eliminating the costs and risks of surgeries. Here, millimeter-scale bioresorbable Fabry-Perot interferometers and two dimensional photonic crystal structures enable precise, continuous measurements of pressure and temperature. Combined mechanical and optical simulations reveal the fundamental sensing mechanisms. In vitro studies and histopathological evaluations quantify the measurement accuracies, operational lifetimes, and biocompatibility of these systems. In vivo demonstrations establish clinically relevant performance attributes. The materials, device designs, and fabrication approaches outlined here establish broad foundational capabilities for diverse classes of bioresorbable optical sensors.

2019 International Consensus on Cardiopulmonary Resuscitation and Emergency Cardiovascular Care Science With Treatment Recommendations: Summary From the Basic Life Support; Advanced Life Support; Pediatric Life Support; Neonatal Life Support; Education, Implementation, and Teams; and First Aid Task Forces

Abstract: The International Liaison Committee on Resuscitation has initiated a continuous review of new, peer-reviewed, published cardiopulmonary resuscitation science. This is the third annual summary of the International Liaison Committee on Resuscitation International Consensus on Cardiopulmonary Resuscitation and Emergency Cardiovascular Care Science With Treatment Recommendations. It addresses the most recent published resuscitation evidence reviewed by International Liaison Committee on Resuscitation Task Force science experts. This summary addresses the role of cardiac arrest centers and dispatcher-assisted cardiopulmonary resuscitation, the role of extracorporeal cardiopulmonary resuscitation in adults and children, vasopressors in adults, advanced airway interventions in adults and children, targeted temperature management in children after cardiac arrest, initial oxygen concentration during resuscitation of newborns, and interventions for presyncope by first aid providers. Members from 6 International Liaison Committee on Resuscitation task forces have assessed, discussed, and debated the certainty of the evidence on the basis of the Grading of Recommendations, Assessment, Development, and Evaluation criteria, and their statements include consensus treatment recommendations. Insights into the deliberations of the task forces are provided in the Justification and Evidence to Decision Framework Highlights sections. The task forces also listed priority knowledge gaps for further research.

New normalized Rortex/vortex identification method

Abstract: A new vortex identification criterion, named ΩR, is proposed for the normalization of Rortex, using the idea of the Omega method (Ω). ΩR is a normalized function from 0 to 1, which measures the relative rotation strength on the plane perpendicular to the local rotation axis. The advantages of the proposed ΩR method can be summarized as follows: (1) ΩR is from 0 to 1 and can be further used in statistics and correlation analysis as a physical quantity. (2) ΩR can distinguish the rotational vortices from the shear layers, discontinuity structures, and other non-physical structures. (3) ΩR is quite robust and can be always set as 0.52 to capture vortex structures in different cases and at different time steps.

Generalized Uncorrelated Regression with Adaptive Graph for Unsupervised Feature Selection

Abstract: Unsupervised feature selection always occupies a key position as a preprocessing in the tasks of classification or clustering due to the existence of extra essential features within high-dimensional data. Although lots of efforts have been made, the existing methods neglect to consider the redundancy of features, and thus select redundant features. In this brief, by virtue of a generalized uncorrelated constraint, we present an improved sparse regression model [generalized uncorrelated regression model (GURM)] for seeking the uncorrelated yet discriminative features. Benefited from this, the structure of data is kept in the Stiefel manifold, which avoids the potential trivial solution triggered by a conventional ridge regression model. Besides that, the uncorrelated constraint equips the model with the closed-form solution. In addition, we also incorporate a graph regularization term based on the principle of maximum entropy into the GURM model (URAFS), so as to embed the local geometric structure of data into the manifold learning. An efficient algorithm is designed to perform URAFS by virtue of the existing generalized powered iteration method. Extensive experiments on eight benchmark data sets among seven state-of-the-art methods on the task of clustering are conducted to verify the effectiveness and superiority of the proposed method.

The Long‐Term Impact of Entrepreneurial Self‐Efficacy and Entrepreneurial Orientation on Venture Performance

Abstract: We collected multi‐wave survey data to assess the lagged effects of entrepreneurial self‐efficacy (ESE) and entrepreneurial orientation (EO) on firm performance over a five‐year period. The results...

Reinforcement Learning-Based Adaptive Optimal Exponential Tracking Control of Linear Systems With Unknown Dynamics

Abstract: Reinforcement learning (RL) has been successfully employed as a powerful tool in designing adaptive optimal controllers. Recently, off-policy learning has emerged to design optimal controllers for systems with completely unknown dynamics. However, current approaches for optimal tracking control design either result in bounded tracking error, rather than zero tracking error, or require partial knowledge of the system dynamics. Moreover, they usually require to collect a large set of data to learn the optimal solution. To obviate these limitations, this paper applies a combination of off-policy learning and experience-replay for output regulation tracking control of continuous-time linear systems with completely unknown dynamics. To this end, the off-policy integral RL-based technique is first used to obtain the optimal control feedback gain, and to explicitly identify the involved system dynamics using the same data. Second, a data-efficient-based experience replay method is developed to compute the exosystem dynamics. Finally, the output regulator equations are solved using data measured online. It is shown that the proposed control method stabilizes the closed-loop tracking error dynamics, and gives an explicit exponential convergence rate for the output tracking error. Simulation results show the effectiveness of the proposed approach.

GrandSLAm: Guaranteeing SLAs for Jobs in Microservices Execution Frameworks

Abstract: The microservice architecture has dramatically reduced user effort in adopting and maintaining servers by providing a catalog of functions as services that can be used as building blocks to construct applications. This has enabled datacenter operators to look at managing datacenter hosting microservices quite differently from traditional infrastructures. Such a paradigm shift calls for a need to rethink resource management strategies employed in such execution environments. We observe that the visibility enabled by a microservices execution framework can be exploited to achieve high throughput and resource utilization while still meeting Service Level Agreements, especially in multi-tenant execution scenarios. In this study, we present GrandSLAm, a microservice execution framework that improves utilization of datacenters hosting microservices. GrandSLAm estimates time of completion of requests propagating through individual microservice stages within an application. It then leverages this estimate to drive a runtime system that dynamically batches and reorders requests at each microservice in a manner where individual jobs meet their respective target latency while achieving high throughput. GrandSLAm significantly increases throughput by up to 3x compared to the our baseline, without violating SLAs for a wide range of real-world AI and ML applications.