arXiv subjects

Nan Liu

Publications and source records attributed to Nan Liu.

At least 145 records · Page 8Linked to original sources

arXiv2021

Deep learning for temporal data representation in electronic health records: A systematic review of challenges and methodologies

Objective: Temporal electronic health records (EHRs) can be a wealth of information for secondary uses, such as clinical events prediction or chronic disease management. However, challenges exist for temporal data representation. We therefore sought to identify these challenges and evaluate novel methodologies for addressing them through a systematic examination of deep learning solutions. Methods: We searched five databases (PubMed, EMBASE, the Institute of Electrical and Electronics Engineers [IEEE] Xplore Digital Library, the Association for Computing Machinery [ACM] digital library, and Web of Science) complemented with hand-searching in several prestigious computer science conference proceedings. We sought articles that reported deep learning methodologies on temporal data representation in structured EHR data from January 1, 2010, to August 30, 2020. We summarized and analyzed the selected articles from three perspectives: nature of time series, methodology, and model implementation. Results: We included 98 articles related to temporal data representation using deep learning. Four major challenges were identified, including data irregularity, data heterogeneity, data sparsity, and model opacity. We then studied how deep learning techniques were applied to address these challenges. Finally, we discuss some open challenges arising from deep learning. Conclusion: Temporal EHR data present several major challenges for clinical prediction modeling and data utilization. To some extent, current deep learning solutions can address these challenges. Future studies can consider designing comprehensive and integrated solutions. Moreover, researchers should incorporate additional clinical domain knowledge into study designs and enhance the interpretability of the model to facilitate its implementation in clinical practice.

cs.LG↗

arXiv2021

AutoScore-Imbalance: An interpretable machine learning tool for development of clinical scores with rare events data

Background: Medical decision-making impacts both individual and public health. Clinical scores are commonly used among a wide variety of decision-making models for determining the degree of disease deterioration at the bedside. AutoScore was proposed as a useful clinical score generator based on machine learning and a generalized linear model. Its current framework, however, still leaves room for improvement when addressing unbalanced data of rare events. Methods: Using machine intelligence approaches, we developed AutoScore-Imbalance, which comprises three components: training dataset optimization, sample weight optimization, and adjusted AutoScore. All scoring models were evaluated on the basis of their area under the curve (AUC) in the receiver operating characteristic analysis and balanced accuracy (i.e., mean value of sensitivity and specificity). By utilizing a publicly accessible dataset from Beth Israel Deaconess Medical Center, we assessed the proposed model and baseline approaches in the prediction of inpatient mortality. Results: AutoScore-Imbalance outperformed baselines in terms of AUC and balanced accuracy. The nine-variable AutoScore-Imbalance sub-model achieved the highest AUC of 0.786 (0.732-0.839) while the eleven-variable original AutoScore obtained an AUC of 0.723 (0.663-0.783), and the logistic regression with 21 variables obtained an AUC of 0.743 (0.685-0.800). The AutoScore-Imbalance sub-model (using down-sampling algorithm) yielded an AUC of 0. 0.771 (0.718-0.823) with only five variables, demonstrating a good balance between performance and variable sparsity. Conclusions: The AutoScore-Imbalance tool has the potential to be applied to highly unbalanced datasets to gain further insight into rare medical events and to facilitate real-world clinical decision-making.

cs.LG↗

arXiv2021

AutoScore-Survival: Developing interpretable machine learning-based time-to-event scores with right-censored survival data

Scoring systems are highly interpretable and widely used to evaluate time-to-event outcomes in healthcare research. However, existing time-to-event scores are predominantly created ad-hoc using a few manually selected variables based on clinician's knowledge, suggesting an unmet need for a robust and efficient generic score-generating method. AutoScore was previously developed as an interpretable machine learning score generator, integrated both machine learning and point-based scores in the strong discriminability and accessibility. We have further extended it to time-to-event data and developed AutoScore-Survival, for automatically generating time-to-event scores with right-censored survival data. Random survival forest provides an efficient solution for selecting variables, and Cox regression was used for score weighting. We illustrated our method in a real-life study of 90-day mortality of patients in intensive care units and compared its performance with survival models (i.e., Cox) and the random survival forest. The AutoScore-Survival-derived scoring model was more parsimonious than survival models built using traditional variable selection methods (e.g., penalized likelihood approach and stepwise variable selection), and its performance was comparable to survival models using the same set of variables. Although AutoScore-Survival achieved a comparable integrated area under the curve of 0.782 (95% CI: 0.767-0.794), the integer-valued time-to-event scores generated are favorable in clinical applications because they are easier to compute and interpret. Our proposed AutoScore-Survival provides an automated, robust and easy-to-use machine learning-based clinical score generator to studies of time-to-event outcomes. It provides a systematic guideline to facilitate the future development of time-to-event scores for clinical applications.

cs.LG↗

arXiv2021

TEM analyses of unusual presolar silicon carbide: Insights into the range of circumstellar dust condensation conditions

Presolar silicon carbide (SiC) grains in meteoritic samples can help constrain circumstellar condensation processes and conditions in C-rich stars and core-collapse supernovae. This study presents our findings on eight presolar SiC grains from AGB stars (four mainstream and one Y grain) and core-collapse supernovae (three X grains), chosen on the basis of μ-Raman spectral features that were indicative of their having unusual non-3C polytypes and/or high degrees of crystal disorder. Analytical transmission electron microscopy (TEM), which provides elemental compositional and structural information, shows evidence for complex histories for the grains. Our TEM results confirm the presence of non-3C,2H crystal domains. Minor element heterogeneities and/or subgrains were observed in all grains analyzed for their compositions. The C/O ratios inferred for the parent stars varied from 0.98 to greater than or equal to 1.03. Our data show that SiC condensation can occur under a wide range of conditions, in which environmental factors other than temperature (e.g., pressure, gas composition, heterogeneous nucleation on pre-condensed phases) play a significant role. Based on previous μ-Raman studies, about 10% of SiC grains may have infrared (IR) spectral features that are influenced by crystal defects, porosity, and/or subgrains. Future sub-diffraction limited IR measurements of complex SiC grains might shed further light on the relative contributions of each of these features to the shape and position of the characteristic IR 11-μm SiC feature and thus improve the interpretation of IR spectra of AGB stars like those that produced the presolar SiC grains.

astro-ph.SR↗

arXiv2020

Cluster analysis of presolar silicon carbide grains: evaluation of their classification and astrophysical implications

Cluster analysis of presolar silicon carbide grains based on literature data for 12C/13C, 14N/15N, δ30Si/28Si, and δ29Si/28Si including or not inferred initial 26Al/27Al data, reveals nine clusters agreeing with previously defined grain types but also highlighting new divisions. Mainstream grains reside in three clusters probably representing different parent star metallicities. One of these clusters has a compact core, with a narrow range of composition, pointing to an enhanced production of SiC grains in asymptotic giant branch (AGB) stars with a narrow range of masses and metallicities. The addition of 26Al/27Al data highlights a cluster of mainstream grains, enriched in 15N and 26Al, which cannot be explained by current AGB models. We defined two AB grain clusters, one with 15N and 26Al excesses, and the other with 14N and smaller 26Al excesses, in agreement with recent studies. Their definition does not use the solar N isotopic ratio as a divider, and the contour of the 26Al-rich AB cluster identified in this study is in better agreement with core-collapse supernova models. We also found a cluster with a mixture of putative nova and AB grains, which may have formed in supernova or nova environments. X grains make up two clusters, having either strongly correlated Si isotopic ratios or deviating from the 2/3 slope line in the Si 3-isotope plot. Finally, most Y and Z grains are jointly clustered, suggesting that the previous use of 12C/13C= 100 as a divider for Y grains was arbitrary. Our results show that cluster analysis is a powerful tool to interpret the data in light of stellar evolution and nucleosynthesis modelling and highlight the need of more multi-element isotopic data for better classification.

astro-ph.SR↗

arXiv2020

The Capacity of Symmetric Private Information Retrieval under Arbitrary Collusion and Eavesdropping Patterns

We study the symmetric private information retrieval (SPIR) problem under arbitrary collusion and eavesdropping patterns for replicated databases. We find its capacity, which is the same as the capacity of the original SPIR problem with the number of databases $N$ replaced by a number $F^*$. The number $F^*$ is the optimal solution to a linear programming problem that is a function of the joint pattern, which is the union of the collusion and eavesdropping pattern. This is the first result that shows how two arbitrary patterns collectively affect the capacity of the PIR problem. We draw the conclusion that for SPIR problems, the collusion and eavesdropping constraints are interchangeable in terms of capacity. As special cases of our result, the capacity of the SPIR problem under arbitrary collusion patterns and the capacity of the PIR problem under arbitrary eavesdropping patterns are also found.

cs.IT↗

arXiv2020

On the mass and metallicity distribution of parent AGB stars of presolar SiC

The vast majority (>=90%) of presolar SiC grains identified in primitive meteorites are relics of ancient asymptotic giant branch (AGB) stars, whose ejecta were incorporated into the Solar System during its formation. Detailed characterization of these ancient stardust grains has revealed precious information on mixing processes in AGB interiors in great detail. However, the mass and metallicity distribution of their parent stars still remains ambiguous, although such information is crucial to investigating the slow neutron capture process, whose efficiency is mass- and metallicity-dependent. Using a well-known Milky Way chemo-dynamical model, we follow the evolution of the AGB stars that polluted the Solar System at 4.57 Gyr ago and weighted the stars based on their SiC dust productions. We find that presolar SiC in the Solar System predominantly originated from AGB stars with M~2 Msun and Z~Zsun. Our finding well explains the grain-size distribution of presolar SiC identified in situ in primitive meteorites. Moreover, it provides complementary results to very recent papers dealing with the characterization of parent stars of presolar SiC.

astro-ph.SR↗

arXiv2020

Magnetic-buoyancy Induced Mixing in AGB Stars: Presolar SiC Grains

Isotope ratios can be measured in presolar SiC grains from ancient Asymptotic Giant Branch (AGB) stars at permil-level (0.1\%) precision. Such precise grain data permit derivation of more stringent constraints and calibrations on mixing efficiency in AGB models than traditional spectroscopic observations. In this paper we compare SiC heavy-element isotope ratios to a new series of FRUITY models that include the effects of mixing triggered by magnetic fields. Based on 2D and 3D simulations available in the literature, we propose a new formulation, upon which the general features of mixing induced by magnetic fields can be derived. The efficiency of such a mixing, on the other hand, relies on physical quantities whose values are poorly constrained. We present here our calibration by comparing our model results with the heavy-element isotope data of presolar SiC grains from AGB stars. We demonstrate that the isotopic compositions of all measured elements (Ni, Sr, Zr, Mo, Ba) can be simultaneously fitted by adopting a single magnetic field configuration in our new FRUITY models.

astro-ph.SR↗

arXiv2020

The anomalous Hall Effect in a Magnetically Extended Topological Insulator Heterostructure

Constructing heterostructures of a topological insulator (TI) with an undoped magnetic insulator (MI) is a clean and versatile approach to break the time-reversible symmetry in the TI surface states. Despite a lot of efforts, the strength of interfacial magnetic proximity effect (MPE) is still too weak to achieve the quantum anomalous Hall effect and many other topological quantum phenomena. Recently, a new approach based on intercalation of atomic layers of MI, referred to as magnetic extension, was proposed to realize strong MPE [2D Mater. 4, 025082(2017)]. Motivated by this proposal, here, we study a magnetic extension system prepared by molecular beam epitaxy growth of MnSe thin films on topological insulator (Bi,Sb)2Te3. Direct evidence is obtained for intercalation of MnSe atomic layer into a few quintuple layers of (Bi,Sb)2Te3, forming either a double magnetic septuple layer (SL) or an isolated single SL at the interface, where one SL denotes a van der Waals building block consisting of B-A-B-Mn-B-A-B (A=Bi1-xSbx, B= Te1-ySey). The two types of interfaces (namely TI/mono-SL and TI/bi-SL) have different MPE, which is manifested as distinctively different transport behaviors. Specifically, the mono-SL induces a spinflip transition with a sharp change at small magnetic field in the anomalous Hall effect of TI layers, while the bi-SL induces a spin-flop transition with a slow change at large field. Our work demonstrates a useful platform to realize the full potential of the magnetic extension approach for pursuing novel topological physics and related device applications.

physics.app-ph↗

arXiv2020

A Blockchain-based Decentralized Federated Learning Framework with Committee Consensus

Federated learning has been widely studied and applied to various scenarios. In mobile computing scenarios, federated learning protects users from exposing their private data, while cooperatively training the global model for a variety of real-world applications. However, the security of federated learning is increasingly being questioned, due to the malicious clients or central servers' constant attack to the global model or user privacy data. To address these security issues, we proposed a decentralized federated learning framework based on blockchain, i.e., a Blockchain-based Federated Learning framework with Committee consensus (BFLC). The framework uses blockchain for the global model storage and the local model update exchange. To enable the proposed BFLC, we also devised an innovative committee consensus mechanism, which can effectively reduce the amount of consensus computing and reduce malicious attacks. We then discussed the scalability of BFLC, including theoretical security, storage optimization, and incentives. Finally, we performed experiments using real-world datasets to verify the effectiveness of the BFLC framework.

cs.DC↗

arXiv2020

The Capacity of Private Information Retrieval Under Arbitrary Collusion Patterns

We study the private information retrieval (PIR) problem under arbitrary collusion pattern for replicated databases. We find its capacity, which is the same as the capacity of the original PIR problem with the number of databases $N$ replaced by a number $S^*$. The number $S^*$ is the optimal solution to a linear programming problem that is a function of the collusion pattern. Hence, the collusion pattern affects the capacity of the PIR problem only through the number $S^*$.

cs.IT↗

arXiv2019

Long-time Asymptotic Behavior of the Fifth-order Modified KdV Equation in Low Regularity Spaces

Based on the nonlinear steepest descent method of Deift and Zhou for oscillatory Riemann--Hilbert problems and the Dbar approach, the long-time asymptotic behavior of solutions to the fifth-order modified Korteweg-de Vries equation on the line is studied in the case of initial conditions that belong to some weighted Sobolev spaces. Using techniques in Fourier analysis and the idea of $I$-method, we give its global well-posedness in lower regularity Sobolev spaces, and then obtain the asymptotic behavior in these spaces with weights.

math.AP↗

arXiv2019

Long-time asymptotic behavior for an extended modified Korteweg-de Vries equation

We investigate an integrable extended modified Korteweg-de Vries equation on the line with the initial value belonging to the Schwartz space. By performing the nonlinear steepest descent analysis of an associated matrix Riemann--Hilbert problem, we obtain the explicit leading-order asymptotics of the solution of this initial value problem as time $t$ goes to infinity. For a special case $α=0$, we present the asymptotic formula of the solution to the extended modified Korteweg-de Vries equation in region $\mathcal{P}=\{(x,t)\in\bfR^2|0<x\leq Mt^{\frac{1}{5}},t\geq3\}$ in terms of the solution of a fourth order Painlevé II equation.

math.AP↗

arXiv2019

Presolar Silicon Carbide Grains of Types Y and Z: Their Molybdenum Isotopic Compositions and Stellar Origins

We report Mo isotopic compositions of 37 presolar SiC grains of types Y (19) and Z (18), rare types commonly argued to have formed in lower-than-solar metallicity asymptotic giant branch (AGB) stars. Direct comparison of the Y and Z grain data with data for mainstream grains from AGB stars of close-to-solar metallicity demonstrates that the three types of grains have indistinguishable Mo isotopic compositions. We show that the Mo isotope data can be used to constrain the maximum stellar temperatures (TMAX) during thermal pulses in AGB stars. Comparison of FRUITY Torino AGB nucleosynthesis model calculations with the grain data for Mo isotopes points to an origin from low-mass (~1.5-3 Msun) rather than intermediate-mass (>3-~9 Msun) AGB stars. Because of the low efficiency of 22Ne(α,n)25Mg at the low TMAX values attained in low-mass AGB stars, model calculations cannot explain the large 30Si excesses of Z grains as arising from neutron capture, so these excesses remain a puzzle at the moment.

astro-ph.SR↗

arXiv2019

Time-sync Video Tag Extraction Using Semantic Association Graph

Time-sync comments reveal a new way of extracting the online video tags. However, such time-sync comments have lots of noises due to users' diverse comments, introducing great challenges for accurate and fast video tag extractions. In this paper, we propose an unsupervised video tag extraction algorithm named Semantic Weight-Inverse Document Frequency (SW-IDF). Specifically, we first generate corresponding semantic association graph (SAG) using semantic similarities and timestamps of the time-sync comments. Second, we propose two graph cluster algorithms, i.e., dialogue-based algorithm and topic center-based algorithm, to deal with the videos with different density of comments. Third, we design a graph iteration algorithm to assign the weight to each comment based on the degrees of the clustered subgraphs, which can differentiate the meaningful comments from the noises. Finally, we gain the weight of each word by combining Semantic Weight (SW) and Inverse Document Frequency (IDF). In this way, the video tags are extracted automatically in an unsupervised way. Extensive experiments have shown that SW-IDF (dialogue-based algorithm) achieves 0.4210 F1-score and 0.4932 MAP (Mean Average Precision) in high-density comments, 0.4267 F1-score and 0.3623 MAP in low-density comments; while SW-IDF (topic center-based algorithm) achieves 0.4444 F1-score and 0.5122 MAP in high-density comments, 0.4207 F1-score and 0.3522 MAP in low-density comments. It has a better performance than the state-of-the-art unsupervised algorithms in both F1-score and MAP.

cs.IR↗

arXiv2019

The Capacity of Multi-round Private Information Retrieval from Byzantine Databases

In this work, we investigate the capacity of private information retrieval (PIR) from $N$ replicated databases, where a subset of the databases are untrustworthy (byzantine) in their answers to the query of the user. We allow for multi-round queries and demonstrate that the identities of the byzantine databases can be determined with a small additional download cost. As a result, the capacity of the multi-round PIR with byzantine databases (BPIR) reaches that of the robust PIR problem when the number of byzantine databases is less than the number of trustworthy databases.

cs.IT↗

arXiv2019

Coded Caching with Heterogeneous Cache Sizes and Link Qualities: The Two-User Case

Centralized coded caching problem is studied for the two-user scenario, considering heterogeneous cache capacities at the users and private channels from the server to the users, in addition to a shared channel. Optimal caching and delivery strategies that minimize the worst-case delivery latency are presented for an arbitrary number of files. The converse proof follows from the sufficiency of file-index-symmetric caching and delivery codes, while the achievability is obtained through memory-sharing among a number of special memory capacity pairs. The optimal scheme is shown to exploit the private link capacities by transmitting part of the corresponding user`s request in an uncoded fashion. When there are no private links, the results presented here improve upon the two known results in the literature, namely, i) equal cache capacities and arbitrary number of files; and ii) unequal cache capacities and $N=2$ files. The results are then extended to the caching problem with heterogeneous distortion requirements.

cs.IT↗

arXiv2018

The Gerdjikov-Ivanov type derivative nonlinear Schrödinger equation: Long-time dynamics of nonzero boundary conditions

We consider the Gerdjikov--Ivanov type derivative nonlinear Schrödinger equation \berr \ii q_{t}+q_{xx}-\ii q^2\bar{q}_{x}+\frac{1}{2}(|q|^4-q_0^4)q=0 \eerr on the line. The initial value $q(x,0)$ is given and satisfies the symmetric, nonzero boundary conditions at infinity, that is, $q(x,0)\rightarrow q_\pm$ as $x\rightarrow\pm\infty$, and $|q_\pm|=q_0>0$. The goal of this paper is to study the asymptotic behavior of the solution of this initial-value problem as $t\rightarrow\infty$. The main tool is the asymptotic analysis of an associated matrix Riemann--Hilbert problem by using the steepest descent method and the so-called $g$-function mechanism. We show that the solution $q(x,t)$ of this initial value problem has a different asymptotic behavior in different regions of the $xt$-plane. In the regions $x<-2\sqrt{2}q_0^2t$ and $x>2\sqrt{2}q_0^2t$, the solution takes the form of a plane wave. In the region $-2\sqrt{2}q_0^2t<x<2\sqrt{2}q_0^2t$, the solution takes the form of a modulated elliptic wave.

math.AP↗