arXiv subjects

Kai Sun

Publications and source records attributed to Kai Sun.

At least 307 records · Page 17Linked to original sources

arXiv2019

Investigating Prior Knowledge for Challenging Chinese Machine Reading Comprehension

Machine reading comprehension tasks require a machine reader to answer questions relevant to the given document. In this paper, we present the first free-form multiple-Choice Chinese machine reading Comprehension dataset (C^3), containing 13,369 documents (dialogues or more formally written mixed-genre texts) and their associated 19,577 multiple-choice free-form questions collected from Chinese-as-a-second-language examinations. We present a comprehensive analysis of the prior knowledge (i.e., linguistic, domain-specific, and general world knowledge) needed for these real-world problems. We implement rule-based and popular neural methods and find that there is still a significant performance gap between the best performing model (68.5%) and human readers (96.0%), especially on problems that require prior knowledge. We further study the effects of distractor plausibility and data augmentation based on translated relevant datasets for English on model performance. We expect C^3 to present great challenges to existing systems as answering 86.8% of questions requires both knowledge within and beyond the accompanying document, and we hope that C^3 can serve as a platform to study how to leverage various kinds of prior knowledge to better understand a given written or orally oriented text. C^3 is available at https://dataset.org/c3/.

cs.CL↗

arXiv2019

Mitigation of Cascading Outages Using a Dynamic Interaction Graph-Based Optimal Power Flow Model

Online decision support for effective mitigation actions against propagating cascading outages in a power grid still poses a big challenge to grid operators. This paper proposes an online mitigation strategy against cascading outages using an optimal power flow model based on a Dynamic Interaction Graph. For a given power grid, its interaction graph, also called an interaction model, was proposed in recent literature, which is composed of components and links that have large contributions to outage propagation. Differing from a conventional interaction graph, a Dynamic Interaction Graph can adaptively update its parameters with changes of grid topology and hence is more accurate in online identification of the most vulnerable transmission lines and likely outage propagation paths for mitigation. Further, the paper introduces an optimal power flow model based on the proposed Dynamic Interaction Graph for determining the optimal control strategy to maintain transfer margins of vulnerable transmission lines and effectively mitigate outage propagation. The numerical results on the IEEE 118-bus system demonstrate the effectiveness of the proposed models and associated control strategy.

eess.SY↗

arXiv2019

Solving Power System Differential Algebraic Equations Using Differential Transformation

This paper proposes a novel non-iterative method to solve power system differential algebraic equations (DAEs) using the differential transformation, a mathematical tool that can obtain power series coefficients by transformation rules instead of calculating high order derivatives and has proved to be effective in solving state variables of nonlinear differential equations in our previous study. This paper further solves non-state variables, e.g. current injections and bus voltages, directly with a realistic DAE model of power grids. These non-state variables, nonlinearly coupled in network equations, are conventionally solved by numerical methods with time-consuming iterations, but their differential transformations are proved to satisfy formally linear equations in this paper. Thus, a non-iterative algorithm is designed to analytically solve all variables of a power system DAE model with ZIP loads. From test results on a Polish 2383-bus system, the proposed method demonstrates fast and reliable time performance compared to traditional numerical methods.

eess.SY↗

arXiv2019

Unsupervised Annotation of Phenotypic Abnormalities via Semantic Latent Representations on Electronic Health Records

The extraction of phenotype information which is naturally contained in electronic health records (EHRs) has been found to be useful in various clinical informatics applications such as disease diagnosis. However, due to imprecise descriptions, lack of gold standards and the demand for efficiency, annotating phenotypic abnormalities on millions of EHR narratives is still challenging. In this work, we propose a novel unsupervised deep learning framework to annotate the phenotypic abnormalities from EHRs via semantic latent representations. The proposed framework takes the advantage of Human Phenotype Ontology (HPO), which is a knowledge base of phenotypic abnormalities, to standardize the annotation results. Experiments have been conducted on 52,722 EHRs from MIMIC-III dataset. Quantitative and qualitative analysis have shown the proposed framework achieves state-of-the-art annotation performance and computational efficiency compared with other methods.

cs.CL↗

arXiv2019

Direct measurement of a nonlocal entangled quantum state

Entanglement and wave function description are two of the core concepts that make quantum mechanics such a unique theory. A method to directly measure the wave function, using Weak Values, was demonstrated by Lundeen et al., Nature \textbf{474}, 188(2011). However, it is not applicable to a scenario of two disjoint systems, where nonlocal entanglement can be a crucial element since that requires obtaining the Weak Values of nonlocal observables. Here, for the first time, we propose a method to directly measure a nonlocal wave function of a bipartite system, using Modular Values. The method is experimentally implemented for a photon pair in a hyper-entangled state, i.e. entangled both in polarization and momentum degrees of freedom.

quant-ph↗

arXiv2019

Robust edge photocurrent response on layered Type II Weyl semimetal WTe2

Photo sensing and energy harvesting based on exotic properties of quantum materials and new operation principles have great potentials to break the fundamental performance limit of conventional photodetectors and solar cells. As topological nontrivial materials, Weyl semimetals have demonstrated novel optoelectronic properties that promise potential applications in photo detection and energy harvesting arising from their gapless linear dispersion near Weyl nodes and Berry field enhanced nonlinear optical effect at the vicinity of Weyl nodes. In this work, we demonstrate robust photocurrent generation from charge separation of photoexctied electron-hole pairs at the edge of Td-WTe2, a type-II Weyl semimetal, due to crystalline-symmetry breaking along certain crystal fracture directions and possibly enhanced by robust fermi-arc type surface states. Using scanning photocurrent microscopy (SPCM) measurements, we further demonstrate that the edge current response is robust over a wide excitation photon energy. We find that this robust feature is highly generic, and shall arise universally in a wide class of quantum materials with similar crystal symmetries. In addition, possible connections between these edge photocurrents and topological properties of Weyl semimetals are explored. The robust and generic edge current response demonstrated in this work provides a new type of charge separation mechanism for photosensing and energy harvesting over broad wavelength range.

cond-mat.mes-hall↗

arXiv2019

Improving Question Answering with External Knowledge

We focus on multiple-choice question answering (QA) tasks in subject areas such as science, where we require both broad background knowledge and the facts from the given subject-area reference corpus. In this work, we explore simple yet effective methods for exploiting two sources of external knowledge for subject-area QA. The first enriches the original subject-area reference corpus with relevant text snippets extracted from an open-domain resource (i.e., Wikipedia) that cover potentially ambiguous concepts in the question and answer options. As in other QA research, the second method simply increases the amount of training data by appending additional in-domain subject-area instances. Experiments on three challenging multiple-choice science QA tasks (i.e., ARC-Easy, ARC-Challenge, and OpenBookQA) demonstrate the effectiveness of our methods: in comparison to the previous state-of-the-art, we obtain absolute gains in accuracy of up to 8.1%, 13.0%, and 12.8%, respectively. While we observe consistent gains when we introduce knowledge from Wikipedia, we find that employing additional QA training instances is not uniformly helpful: performance degrades when the added instances exhibit a higher level of difficulty than the original training data. As one of the first studies on exploiting unstructured external knowledge for subject-area QA, we hope our methods, observations, and discussion of the exposed limitations may shed light on further developments in the area.

cs.CL↗

arXiv2019

Valley-Hall In-Plane Edge States as Building Blocks for Elastodynamic Logic Circuits

In this work, we investigate theoretically and demonstrate experimentally the existence of valley-Hall edge states in the in-plane dynamics of honeycomb lattices with bi-valued strut thickness. We exploit these states to achieve non-trivial waveguiding of optical modes that is immune to backscattering from sharp corners. We also present how different types of interfaces can be combined into multi-branch junctions to form complex waveguide paths and realize a variety of structural logic designs with unconventional wave transport capabilities. We illustrate this potential with two applications. The first is a direction-selective energy-splitting waveguide tree featuring a pronounced asymmetric wave transport behavior. The second is an internal waveguide loop along which the energy can be temporarily trapped and periodically released, effectively working as a signal delayer. The modal complexity of in-plane elasticity has important consequences on the regime of manifestation of the edge states, as the availability of viable total bandgaps is shifted to higher frequencies compared to the out-of-plane counterpart problem. It also poses additional experimental challenges, associated with proper acquisition and deciphering of the in-plane modes, the solution of which requires a systematic use of in-plane laser vibrometry.

physics.app-ph↗

arXiv2019

Improving Pre-Trained Multilingual Models with Vocabulary Expansion

Recently, pre-trained language models have achieved remarkable success in a broad range of natural language processing tasks. However, in multilingual setting, it is extremely resource-consuming to pre-train a deep language model over large-scale corpora for each language. Instead of exhaustively pre-training monolingual language models independently, an alternative solution is to pre-train a powerful multilingual deep language model over large-scale corpora in hundreds of languages. However, the vocabulary size for each language in such a model is relatively small, especially for low-resource languages. This limitation inevitably hinders the performance of these multilingual models on tasks such as sequence labeling, wherein in-depth token-level or sentence-level understanding is essential. In this paper, inspired by previous methods designed for monolingual settings, we investigate two approaches (i.e., joint mapping and mixture mapping) based on a pre-trained multilingual model BERT for addressing the out-of-vocabulary (OOV) problem on a variety of tasks, including part-of-speech tagging, named entity recognition, machine translation quality estimation, and machine reading comprehension. Experimental results show that using mixture mapping is more promising. To the best of our knowledge, this is the first work that attempts to address and discuss the OOV issue in multilingual settings.

cs.CL↗

arXiv2019

Observation of a ferro-rotational order coupled with second-order nonlinear optical fields

The ferro-rotational order, whose order parameter (OP) is an axial vector invariant under both time reversal (TR) and spatial inversion (SI) operations, is the last remaining category of ferroics to be observed after the ferroelectric, ferromagnetic, and ferro-toroidal orders. This order has become increasingly popular in many new quantum materials, especially in complex oxides, and is considered responsible for a number of novel phenomena such as polar vortices, giant magnetoelectric coupling, and type-II multiferroics. However, physical properties of the ferro-rotational order have been rarely studied either theoretically or experimentally. Here, using high sensitivity rotational anisotropy second harmonic generation (RA SHG), we have, for the first time, exploited the electric quadrupole (EQ) contribution to the SHG to directly couple to this centrosymmetric ferro-rotational order in an archetype of type-II multiferroics, RbFe(MoO4)2. Surprisingly, we have found that two types of domains with opposite ferro-rotational vectors emerge with distinct populations at the critical temperature Tc ~195 K and gradually evolve to reach an even ratio at lower temperatures. Moreover, we have identified the ferro-rotational order phase transition as weak first order, and have revealed its conjugate coupling field as a unique combination of the induced EQ SHG and the incident fundamental electric fields. Our results on physical properties of a ferro-rotational order provide crucial knowledge for understanding and searching for novel phases of matter built upon the ferro-rotational order. Further, these results open the possibility of revealing unconventional centrosymmetric orders and identifying their conjugate coupling fields with second order nonlinear optics.

cond-mat.mtrl-sci↗

arXiv2019

Detection of quantum phases via out-of-time-order correlators

We elucidate the relation between out-of-time-order correlators (OTOCs) and quantum phase transitions via analytically studying the OTOC dynamics in a degenerate spectrum. Our method points to key ingredients to dynamically detect quantum phases via out-of-time-order correlators for a wide range of quantum phase transitions and explains the existing numerical results in the literature. We apply our method to a critical model, XXZ model that numerically confirms our predictions.

quant-ph↗

arXiv2019

Evidence Sentence Extraction for Machine Reading Comprehension

Remarkable success has been achieved in the last few years on some limited machine reading comprehension (MRC) tasks. However, it is still difficult to interpret the predictions of existing MRC models. In this paper, we focus on extracting evidence sentences that can explain or support the answers of multiple-choice MRC tasks, where the majority of answer options cannot be directly extracted from reference documents. Due to the lack of ground truth evidence sentence labels in most cases, we apply distant supervision to generate imperfect labels and then use them to train an evidence sentence extractor. To denoise the noisy labels, we apply a recently proposed deep probabilistic logic learning framework to incorporate both sentence-level and cross-sentence linguistic indicators for indirect supervision. We feed the extracted evidence sentences into existing MRC models and evaluate the end-to-end performance on three challenging multiple-choice MRC datasets: MultiRC, RACE, and DREAM, achieving comparable or better performance than the same models that take as input the full reference document. To the best of our knowledge, this is the first work extracting evidence sentences for multiple-choice MRC.

cs.CL↗

arXiv2019

Revealing Fermionic Quantum Criticality from New Monte Carlo Techniques

This review summarizes recent developments in the study of fermionic quantum criticality, focusing on new progress in numerical methodologies, especially quantum Monte Carlo methods, and insights that emerged from recently large-scale numerical simulations. Quantum critical phenomena in fermionic systems have attracted decades of extensive research efforts, partially lured by their exotic properties and potential technology applications and partially awaked by the profound and universal fundamental principles that govern these quantum critical systems. Due to the complex and non-perturbative nature, these systems belong to the most difficult and challenging problems in the study of modern condensed matter physics, and many important fundamental problems remain open. Recently, new developments in model design and algorithm improvements enabled unbiased large-scale numerical solutions to be achieved in the close vicinity of these quantum critical points, which paves a new pathway towards achieving controlled conclusions through combined efforts of theoretical and numerical studies, as well as possible theoretical guidance for experiments in heavy-fermion compounds, Cu-based and Fe-based superconductors, ultra-cold fermionic atomic gas, twisted graphene layers, etc., where signatures of fermionic quantum criticality exist.

cond-mat.str-el↗

arXiv2019

On Distribution Patterns of Power Flow Solutions

A fundamental challenge in computer analysis of power flow is the rigorous understanding of the impact of different loading levels on the solutions of the power flow equation. This letter presents a comprehensive study of possible numerical solutions that may arise as the loading level varies. In particular, a type of "false" load flow solutions is reported for the first time as a legitimate numerical solutions but with no engineering justification. The existence and the mechanism of why this category of solutions exist are rigorously analyzed. The probability mass function of voltage solution is shown to be less dependent on loading levels. Furthermore, numbers of all actual solutions and those solutions within engineering limits are summarized. This letter presents a first attempt to compute such huge numbers of solutions to the tested systems, and analyze their distribution patterns. All solution sets investigated in this letter are posted online associated with this letter to support any further research purposes.

eess.SP↗

arXiv2019

Blockchain-based Personal Data Management: From Fiction to Solution

The emerging blockchain technology has enabled various decentralised applications in a trustless environment without relying on a trusted intermediary. It is expected as a promising solution to tackle sophisticated challenges on personal data management, thanks to its advanced features such as immutability, decentralisation and transparency. Although certain approaches have been proposed to address technical difficulties in personal data management; most of them only provided preliminary methodological exploration. Alarmingly, when utilising Blockchain for developing a personal data management system, fictions have occurred in existing approaches and been promulgated in the literature. Such fictions are theoretically doable; however, by thoroughly breaking down consensus protocols and transaction validation processes, we clarify that such existing approaches are either impractical or highly inefficient due to the natural limitations of the blockchain and Smart Contracts technologies. This encourages us to propose a feasible solution in which such fictions are reduced by designing a novel system architecture with a blockchain-based "proof of permission" protocol. We demonstrate the feasibility and efficiency of the proposed models by implementing a clinical data sharing service built on top of a public blockchain platform. We believe that our research resolves existing ambiguity and take a step further on providing a practically feasible solution for decentralised personal data management.

cs.DC↗

arXiv2019

Differential Transformation of a Motor Load Model for Time-Domain Simulation

The Differential Transformation (DT) method has demonstrated its potential in speeding up power system time-domain simulation by our previous work. This letter further derives DTs about a motor load model and proves that the nonlinear current injection equation about a motor load can be transformed into a linear equation by means of DT.

eess.SY↗

arXiv2019

Integrated Multi-omics Analysis Using Variational Autoencoders: Application to Pan-cancer Classification

Different aspects of a clinical sample can be revealed by multiple types of omics data. Integrated analysis of multi-omics data provides a comprehensive view of patients, which has the potential to facilitate more accurate clinical decision making. However, omics data are normally high dimensional with large number of molecular features and relatively small number of available samples with clinical labels. The "dimensionality curse" makes it challenging to train a machine learning model using high dimensional omics data like DNA methylation and gene expression profiles. Here we propose an end-to-end deep learning model called OmiVAE to extract low dimensional features and classify samples from multi-omics data. OmiVAE combines the basic structure of variational autoencoders with a classification network to achieve task-oriented feature extraction and multi-class classification. The training procedure of OmiVAE is comprised of an unsupervised phase without the classifier and a supervised phase with the classifier. During the unsupervised phase, a hierarchical cluster structure of samples can be automatically formed without the need for labels. And in the supervised phase, OmiVAE achieved an average classification accuracy of 97.49% after 10-fold cross-validation among 33 tumour types and normal samples, which shows better performance than other existing methods. The OmiVAE model learned from multi-omics data outperformed that using only one type of omics data, which indicates that the complementary information from different omics datatypes provides useful insights for biomedical tasks like cancer classification.

cs.LG↗

arXiv2019

Switchable phonon diodes using nonlinear topological Maxwell lattices

Recent progress in topological mechanics have revealed a family of Maxwell lattices that exhibit topologically protected floppy edge modes. These modes lead to a strongly asymmetric elastic wave response. In this paper, we show how topological Maxwell lattices can be used to realize non-reciprocal transmission of elastic waves. Our design leverages the asymmetry associated with the availability of topological floppy edge modes and the geometric nonlinearity built in the mechanical systems response to achieve the desired non-reciprocal behavior, which can be further turned into strongly one-way phonon transport via the addition of on-site pinning potentials. Moreover, we show that the non-reciprocal wave transmission can be switched on and off via topological phase transitions, paving the way to the design of cellular metamaterials that can serve as tunable topologically protected phonon diodes.

cond-mat.soft↗