Searcharxiv⌕ Search

arXiv subjects

Yu Deng

Publications and source records attributed to Yu Deng.

125 records · Page 7Linked to original sources
arXiv2019

Evaluating the Portability of an NLP System for Processing Echocardiograms: A Retrospective, Multi-site Observational Study

While natural language processing (NLP) of unstructured clinical narratives holds the potential for patient care and clinical research, portability of NLP approaches across multiple sites remains a major challenge. This study investigated the portability of an NLP system developed initially at the Department of Veterans Affairs (VA) to extract 27 key cardiac concepts from free-text or semi-structured echocardiograms from three academic medical centers: Weill Cornell Medicine, Mayo Clinic and Northwestern Medicine. While the NLP system showed high precision and recall measurements for four target concepts (aortic valve regurgitation, left atrium size at end systole, mitral valve regurgitation, tricuspid valve regurgitation) across all sites, we found moderate or poor results for the remaining concepts and the NLP system performance varied between individual sites.

cs.CL↗
arXiv2019

Optimal control problems with control complementarity constraints

A special class of optimal control problems with complementarity constraints on the control functions is studied. It is shown that such problems possess optimal solutions whenever the underlying control space is a first-order Sobolev space. After deriving necessary optimality conditions of strong stationarity-type, a penalty method based on the Fischer-Burmeister function is suggested and its theoretical properties are analyzed. Finally, the numerical treatment of the problem is discussed and results of computational experiments are presented.

math.OC↗
arXiv2018

Characterizing Design Patterns of EHR-Driven Phenotype Extraction Algorithms

The automatic development of phenotype algorithms from Electronic Health Record data with machine learning (ML) techniques is of great interest given the current practice is very time-consuming and resource intensive. The extraction of design patterns from phenotype algorithms is essential to understand their rationale and standard, with great potential to automate the development process. In this pilot study, we perform network visualization on the design patterns and their associations with phenotypes and sites. We classify design patterns using the fragments from previously annotated phenotype algorithms as the ground truth. The classification performance is used as a proxy for coherence at the attribution level. The bag-of-words representation with knowledge-based features generated a good performance in the classification task (0.79 macro-f1 scores). Good classification accuracy with simple features demonstrated the attribution coherence and the feasibility of automatic identification of design patterns. Our results point to both the feasibility and challenges of automatic identification of phenotyping design patterns, which would power the automatic development of phenotype algorithms.

cs.CL↗
arXiv2018

Validation of a track repeating algorithm for intensity modulated Carbon therapy with GEANT4

The Fast Dose Calculator (FDC), a track repeating algorithm Monte Carlo method was initially developed for proton therapy. The validation for proton therapy has been demonstrated in a previous work. This method can be expanded to ion applications. Our purpose of this paper is to validate the FDC for carbon therapy. We compare the 3D dose distributions and dose-volume-histograms (DVH) for carbon calculated by FDC with a full Monte Carlo method, GEANT 4. 19 patients in total will be discussed, including 3 patients of prostate, 5 of brain, 3 of head and neck, 4 of lung and 4 of spine. We use gamma-index technique to analyse dose distributions and we do dosimetric analysis for DVH, a more direct and informative quantity for planning system assessment. The FDC calculations of both quantities agree with GEANT4. The gamma-index passing rates of all patients discussed in this paper are above 90% with the criterion 1%/1 mm, above 98% with the criterion 2%/2 mm and over 99.9% with the criterion 3%/3 mm. The Root Mean Square (RMS) of percent difference of dosimetric indices D02, D05, D50, D95 and D98 are 0.75%, 0.70%, 0.79%, 0.83% and 0.76%. And all the difference are allowed for clinical use.

physics.med-ph↗
arXiv2018

Natural Language Processing for EHR-Based Computational Phenotyping

This article reviews recent advances in applying natural language processing (NLP) to Electronic Health Records (EHRs) for computational phenotyping. NLP-based computational phenotyping has numerous applications including diagnosis categorization, novel phenotype discovery, clinical trial screening, pharmacogenomics, drug-drug interaction (DDI) and adverse drug event (ADE) detection, as well as genome-wide and phenome-wide association studies. Significant progress has been made in algorithm development and resource construction for computational phenotyping. Among the surveyed methods, well-designed keyword search and rule-based systems often achieve good performance. However, the construction of keyword and rule lists requires significant manual effort, which is difficult to scale. Supervised machine learning models have been favored because they are capable of acquiring both classification patterns and structures from data. Recently, deep learning and unsupervised learning have received growing attention, with the former favored for its performance and the latter for its ability to find novel phenotypes. Integrating heterogeneous data sources have become increasingly important and have shown promise in improving model performance. Often better performance is achieved by combining multiple modalities of information. Despite these many advances, challenges and opportunities remain for NLP-based computational phenotyping, including better model interpretability and generalizability, and proper characterization of feature relations in clinical narratives

cs.CL↗
arXiv2018

On the global behavior of weak null quasilinear wave equations

We consider a class of quasilinear wave equations in $3+1$ space-time dimensions that satisfy the "weak null condition" as defined by Lindblad and Rodnianski \cite{LR1}, and study the large time behavior of solutions to the Cauchy problem. The prototype for the class of equations considered is $-\partial_t^2 u + (1+u) Δu = 0$. Global solutions for such equations have been constructed by Lindblad \cite{Lindblad1,Lindblad2} and Alinhac \cite{Alinhac1}. Our main results are the derivation of a precise asymptotic system with good error bounds, and a detailed description of the behavior of solutions close to the light cone, including the blow-up at infinity.

math.AP↗
arXiv2018

Long time instability of the Couette flow in low Gevrey spaces

We prove the instability of the Couette flow if the disturbances is less smooth than the Gevrey space of class 2. This shows that this is the critical regularity for this problem since it was proved in [5] that stability and inviscid damping hold for disturbances which are smoother than the Gevrey space of class 2. A big novelty is that this critical space is due to an instability mechanism which is completely nonlinear and is due to some energy cascade.

math.AP↗
arXiv2017

Fast Nearest-Neighbor Classification using RNN in Domains with Large Number of Classes

In scenarios involving text classification where the number of classes is large (in multiples of 10000s) and training samples for each class are few and often verbose, nearest neighbor methods are effective but very slow in computing a similarity score with training samples of every class. On the other hand, machine learning models are fast at runtime but training them adequately is not feasible using few available training samples per class. In this paper, we propose a hybrid approach that cascades 1) a fast but less-accurate recurrent neural network (RNN) model and 2) a slow but more-accurate nearest-neighbor model using bag of syntactic features. Using the cascaded approach, our experiments, performed on data set from IT support services where customer complaint text needs to be classified to return top-$N$ possible error codes, show that the query-time of the slow system is reduced to $1/6^{th}$ while its accuracy is being improved. Our approach outperforms an LSH-based baseline for query-time reduction. We also derive a lower bound on the accuracy of the cascaded model in terms of the accuracies of the individual models. In any two-stage approach, choosing the right number of candidates to pass on to the second stage is crucial. We prove a result that aids in choosing this cutoff number for the cascaded system.

cs.IR↗
arXiv2017

Growth of solutions to NLS on irrational tori

We prove polynomial bounds on the growth of higher Sobolev norms for the nonlinear Schrodinger equation set on a torus, in dimension 3, with super-cubic and sub-quintic nonlinearity. Due to improved Strichartz estimates, these bounds are better for irrational tori than they are for rational tori.

math.AP↗
arXiv2016

The Euler--Maxwell system for electrons: global solutions in $2D$

A basic model for describing plasma dynamics is given by the Euler-Maxwell system, in which compressible ion and electron fluids interact with their own self-consistent electromagnetic field. In this paper we consider the "one-fluid" Euler--Maxwell model for electrons, in 2 spatial dimensions, and prove global stability of a constant neutral background.

math.AP↗
arXiv2016

Multispeed Klein-Gordon systems in dimension three

We consider long time evolution of small solutions to general multispeed Klein-Gordon systems in 3+1 dimensions. We prove that such solutions are always global and scatter to a linear flow, thus extending previous partial results. The main new ingredients of our method is an improved linear dispersion estimate exploiting the asymptotic spherical symmetry of Klein-Gordon waves, and a corresponding bilinear oscillatory integral estimate.

math.AP↗
arXiv2012

Invariance of the Gibbs measure for the Benjamin-Ono equation

In this paper we consider the periodic Benjemin-Ono equation. We will establish the invariance of the Gibbs measure associated to this equation, thus answering a question raised in Tzvetkov [20]. As an intermediate step, we also obtain a local well-posedness result in Besov-type spaces rougher than L^2, extending the L^2 well-posedness result of Molinet [14].

math.AP↗
arXiv2010

Two Dimensional NLS Equation With Random Radial Data

In this paper we study radial solutions of certain two-dimensional nonlinear Schrödinger equation with harmonic potential, which is supercritical with respect to the initial data. By combining the nonlinear smoothing effect of Schr ödinger equation with Lp estimates of Laguerre functions, we are able to prove an almost sure global well-posedness result and the invariance of the Gibbs measure. We also discuss an application to the NLS equation without harmonic potential.

math.AP↗
arXiv2009

OntoELAN: An Ontology-based Linguistic Multimedia Annotator

Despite its scientific, political, and practical value, comprehensive information about human languages, in all their variety and complexity, is not readily obtainable and searchable. One reason is that many language data are collected as audio and video recordings which imposes a challenge to document indexing and retrieval. Annotation of multimedia data provides an opportunity for making the semantics explicit and facilitates the searching of multimedia documents. We have developed OntoELAN, an ontology-based linguistic multimedia annotator that features: (1) support for loading and displaying ontologies specified in OWL; (2) creation of a language profile, which allows a user to choose a subset of terms from an ontology and conveniently rename them if needed; (3) creation of ontological tiers, which can be annotated with profile terms and, therefore, corresponding ontological terms; and (4) saving annotations in the XML format as Multimedia Ontology class instances and, linked to them, class instances of other ontologies used in ontological tiers. To our best knowledge, OntoELAN is the first audio/video annotation tool in linguistic domain that provides support for ontology-based annotation.

cs.DL↗
arXiv2009

ImageSpace: An Environment for Image Ontology Management

More and more researchers have realized that ontologies will play a critical role in the development of the Semantic Web, the next generation Web in which content is not only consumable by humans, but also by software agents. The development of tools to support ontology management including creation, visualization, annotation, database storage, and retrieval is thus extremely important. We have developed ImageSpace, an image ontology creation and annotation tool that features (1) full support for the standard web ontology language DAML+OIL; (2) image ontology creation, visualization, image annotation and display in one integrated framework; (3) ontology consistency assurance; and (4) storing ontologies and annotations in relational databases. It is expected that the availability of such a tool will greatly facilitate the creation of image repositories as islands of the Semantic Web.

cs.DL↗