arXiv subjects

Pengfei Liu

Publications and source records attributed to Pengfei Liu.

At least 253 records · Page 14Linked to original sources

arXiv2019

A van der Waals antiferromagnetic topological insulator with weak interlayer magnetic coupling

Magnetic topological insulators (TI) provide an important material platform to explore quantum phenomena such as quantized anomalous Hall (QAH) effect and Majorana modes, etc. Their successful material realization is thus essential for our fundamental understanding and potential technical revolutions. By realizing a bulk van der Waals material MnBi4Te7 with alternating septuple [MnBi2Te4] and quintuple [Bi2Te3] layers, we show that it is ferromagnetic in plane but antiferromagnetic along the c axis with an out-of-plane saturation field of ~ 0.22 T at 2 K. Our angle-resolved photoemission spectroscopy measurements and first-principles calculations further demonstrate that MnBi4Te7 is a Z2 antiferromagnetic TI with two types of surface states associated with the [MnBi2Te4] or [Bi2Te3] termination, respectively. Additionally, its superlattice nature may make various heterostructures of [MnBi2Te4] and [Bi2Te3] layers possible by exfoliation. Therefore, the low saturation field and the superlattice nature of MnBi4Te7 make it an ideal system to investigate rich emergent phenomena.

cond-mat.mtrl-sci↗

arXiv2019

Gapless surface Dirac cone in antiferromagnetic topological insulator MnBi$_2$Te$_4$

The recent discovered antiferromagnetic topological insulators in Mn-Bi-Te family with intrinsic magnetic ordering have rapidly drawn broad interest since its cleaved surface state is believed to be gapped, hosting the unprecedented axion states with half-integer quantum Hall effect. Here, however, we show unambiguously by using high-resolution angle-resolved photoemission spectroscopy that a gapless Dirac cone at the (0001) surface of MnBi$_2$Te$_4$ exists between the bulk band gap. Such unexpected surface state remains unchanged across the bulk Néel temperature, and is even robust against severe surface degradation, indicating additional topological protection. Through symmetry analysis and $\textit{ab}$-$\textit{initio}$ calculations we consider different types of surface reconstruction of the magnetic moments as possible origins giving rise to such linear dispersion. Our results reveal that the intrinsic magnetic topological insulator hosts a rich platform to realize various topological phases such as topological crystalline insulator and time-reversal-preserved topological insulator, by tuning the magnetic configurations.

cond-mat.mtrl-sci↗

arXiv2019

A Closer Look at Data Bias in Neural Extractive Summarization Models

In this paper, we take stock of the current state of summarization datasets and explore how different factors of datasets influence the generalization behaviour of neural extractive summarization models. Specifically, we first propose several properties of datasets, which matter for the generalization of summarization models. Then we build the connection between priors residing in datasets and model designs, analyzing how different properties of datasets influence the choices of model structure design and training methods. Finally, by taking a typical dataset as an example, we rethink the process of the model design based on the experience of the above analysis. We demonstrate that when we have a deep understanding of the characteristics of datasets, a simple approach can bring significant improvements to the existing state-of-the-art model.A

cs.CL↗

arXiv2019

Exploring Domain Shift in Extractive Text Summarization

Although domain shift has been well explored in many NLP applications, it still has received little attention in the domain of extractive text summarization. As a result, the model is under-utilizing the nature of the training data due to ignoring the difference in the distribution of training sets and shows poor generalization on the unseen domain. With the above limitation in mind, in this paper, we first extend the conventional definition of the domain from categories into data sources for the text summarization task. Then we re-purpose a multi-domain summarization dataset and verify how the gap between different domains influences the performance of neural summarization models. Furthermore, we investigate four learning strategies and examine their abilities to deal with the domain shift problem. Experimental results on three different settings show their different characteristics in our new testbed. Our source code including \textit{BERT-based}, \textit{meta-learning} methods for multi-domain summarization learning and the re-purposed dataset \textsc{Multi-SUM} will be available on our project: \url{http://pfliu.com/TransferSum/}.

cs.CL↗

arXiv2019

Zero-shot Text-to-SQL Learning with Auxiliary Task

Recent years have seen great success in the use of neural seq2seq models on the text-to-SQL task. However, little work has paid attention to how these models generalize to realistic unseen data, which naturally raises a question: does this impressive performance signify a perfect generalization model, or are there still some limitations? In this paper, we first diagnose the bottleneck of text-to-SQL task by providing a new testbed, in which we observe that existing models present poor generalization ability on rarely-seen data. The above analysis encourages us to design a simple but effective auxiliary task, which serves as a supportive model as well as a regularization term to the generation task to increase the models generalization. Experimentally, We evaluate our models on a large text-to-SQL dataset WikiSQL. Compared to a strong baseline coarse-to-fine model, our models improve over the baseline by more than 3% absolute in accuracy on the whole dataset. More interestingly, on a zero-shot subset test of WikiSQL, our models achieve 5% absolute accuracy gain over the baseline, clearly demonstrating its superior generalizability.

cs.CL↗

arXiv2019

Non-trivial topology in a layered Dirac nodal-line semimetal candidate SrZnSb$_2$ with distorted Sb square nets

Dirac states hosted by Sb/Bi square nets are known to exist in the layered antiferromagnetic AMnX$_2$ (A = Ca/Sr/Ba/Eu/Yb, X=Sb/Bi) material family the space group to be P4/nmm or I4/mmm. In this paper, we present a comprehensive study of quantum transport behaviors, angle-resolved photoemission spectroscopy (ARPES) and first-principles calculations on SrZnSb2, a nonmagnetic analogue to AMnX2, which crystallizes in the pnma space group with distorted square nets. From the quantum oscillation measurements up to 35 T, three major frequencies including F$_1$ = 103 T, F$_2$ = 127 T and F$_3$ = 160 T, are identified. The effective masses of the quasiparticles associated with these frequencies are extracted, namely, m*$_1$ = 0.1 m$_e$, m*$_2$ = 0.1 m$_e$ and m*$_3$ = 0.09m$_e$, where m$_e$ is the free electron mass. From the three-band Lifshitz-Kosevich fit, the Berry phases accumulated along the cyclotron orbit of the quasiparticles are 0.06$π$, 1.2$π$ and 0.74$π$ for F$_1$, F$_2$ and F$_3$, respectively. Combined with the ARPES data and the first-principles calculations, we reveal that F2 and F3 are associated with the two nontrivial Fermi pockets at the Brillouin zone edge while F1 is associated with the trivial Fermi pocket at the zone center. In addition, the first-principles calculations further suggest the existence of Dirac nodal line in the band structure of SrZnSb$_2$.

cond-mat.mtrl-sci↗

arXiv2019

DropAttention: A Regularization Method for Fully-Connected Self-Attention Networks

Variants dropout methods have been designed for the fully-connected layer, convolutional layer and recurrent layer in neural networks, and shown to be effective to avoid overfitting. As an appealing alternative to recurrent and convolutional layers, the fully-connected self-attention layer surprisingly lacks a specific dropout method. This paper explores the possibility of regularizing the attention weights in Transformers to prevent different contextualized feature vectors from co-adaption. Experiments on a wide range of tasks show that DropAttention can improve performance and reduce overfitting.

cs.CL↗

arXiv2019

Searching for Effective Neural Extractive Summarization: What Works and What's Next

The recent years have seen remarkable success in the use of deep neural networks on text summarization. However, there is no clear understanding of \textit{why} they perform so well, or \textit{how} they might be improved. In this paper, we seek to better understand how neural extractive summarization systems could benefit from different types of model architectures, transferable knowledge and learning schemas. Additionally, we find an effective way to improve current frameworks and achieve the state-of-the-art result on CNN/DailyMail by a large margin based on our observations and analyses. Hopefully, our work could provide more clues for future research on extractive summarization.

cs.CL↗

arXiv2019

Rational design principles of quantum anomalous Hall effect from superlattice-like magnetic topological insulators

As one of paradigmatic phenomena in condensed matter physics, the quantum anomalous Hall effect (QAHE) in stoichiometric Chern insulators has drawn great interest for years. By using model Hamiltonian analysis and first-principle calculations, we establish a topological phase diagram and map on it with different two-dimensional configurations, which is taken from the recently-grown magnetic topological insulators MnBi4Te7 and MnBi6Te10 with superlattice-like stacking patterns. These configurations manifest various topological phases, including quantum spin Hall effect with and without time-reversal symmetry, as well as QAHE. We then provide design principles to trigger QAHE by tuning experimentally accessible knobs, such as slab thickness and magnetization. Our work reveals that superlattice-like magnetic topological insulators with tunable exchange interaction serve as an ideal platform to realize the long-sought QAHE in pristine compounds, paving a new avenue within the area of topological materials.

cond-mat.mes-hall↗

arXiv2019

TIGS: An Inference Algorithm for Text Infilling with Gradient Search

Text infilling is defined as a task for filling in the missing part of a sentence or paragraph, which is suitable for many real-world natural language generation scenarios. However, given a well-trained sequential generative model, generating missing symbols conditioned on the context is challenging for existing greedy approximate inference algorithms. In this paper, we propose an iterative inference algorithm based on gradient search, which is the first inference algorithm that can be broadly applied to any neural sequence generative models for text infilling tasks. We compare the proposed method with strong baselines on three text infilling tasks with various mask ratios and different mask strategies. The results show that our proposed method is effective and efficient for fill-in-the-blank tasks, consistently outperforming all baselines.

cs.CL↗

arXiv2019

Symmetry-assisted protection and compensation of hidden spin polarization in centrosymmetric systems

It is recently noted that in certain centrosymmetric compounds, spin-orbit interaction couples each local sector that lacks inversion symmetry and thus leads to visible spin polarization effects in the real space, dubbed as 'hidden spin polarization (HSP)'. However, observable spin polarization of a given local sector suffers interference by its inversion partner, impeding material realization and potential applications of HSP. Starting from a single-orbital tight-binding model, we propose a nontrivial way to protect strong sector-projected spin texture through minimizing the interaction between inversion partners by nonsymmorphic symmetry. The HSP effect is generally compensated by inversion partners near the Γ point but immune from interaction around the boundary of the Brillouin zone. We further summarize 17 layer groups that support such symmetry-assisted HSP and identify by first-principle calculations hundreds of quasi-2D materials from the existing databases, among which a group of rare-earth compounds LnOI (Ln = Pr, Nd, Ho, Tm and Lu) serves as great candidates showing strong Rashba- and Dresselhaus-type HSP. Our finding also provides an ideal platform to control HSP for emergent physical effects, such as opening a hybridization gap by tensile strain to realize the time-reversal-invariant topological superconductivity.

cond-mat.mes-hall↗

arXiv2018

Drug cell line interaction prediction

Understanding the phenotypic drug response on cancer cell lines plays a vital rule in anti-cancer drug discovery and re-purposing. The Genomics of Drug Sensitivity in Cancer (GDSC) database provides open data for researchers in phenotypic screening to test their models and methods. Previously, most research in these areas starts from the fingerprints or features of drugs, instead of their structures. In this paper, we introduce a model for phenotypic screening, which is called twin Convolutional Neural Network for drugs in SMILES format (tCNNS). tCNNS is comprised of CNN input channels for drugs in SMILES format and cancer cell lines respectively. Our model achieves $0.84$ for the coefficient of determinant($R^2$) and $0.92$ for Pearson correlation($R_p$), which are significantly better than previous works\cite{ammad2014integrative,haider2015copula,menden2013machine}. Besides these statistical metrics, tCNNS also provides some insights into phenotypic screening.

q-bio.QM↗

arXiv2018

Two-dimensional spin-valley-coupled Dirac semimetals in functionalized SbAs monolayers

In the presence of spin-orbit coupling (SOC), achieving both spin and valley polarized Dirac state is significant to promote the fantastic integration of Dirac physics, spintronics and valleytronics. Based on ab initio calculations, here we demonstrate that a class of spin-valley-coupled Dirac semimetals (svc-DSMs) in the functionalized SbAs monolayers (MLs) can host such desired state. Distinguished from the graphene-like 2D Dirac materials, the Dirac cones in svc-DSMs hold giant spin-splitting induced by strong SOC under inversion symmetry breaking. In the 2.3% strained SbAsH2 ML, the Dirac fermions in inequivalent valleys have opposite Berry curvature and spin moment, giving rise to Dirac spin-valley Hall effect with constant spin Hall conductivity as well as massless and dissipationless transport. Topological analysis reveals that the svc-DSM emerges at the boundary between trivial and 2D topological insulators, which provides a promising platform for realizing the flexible and controllable tuning among different quantum states.

physics.comp-ph↗

arXiv2018

Multi-task Learning over Graph Structures

We present two architectures for multi-task learning with neural sequence models. Our approach allows the relationships between different tasks to be learned dynamically, rather than using an ad-hoc pre-defined structure as in previous work. We adopt the idea from message-passing graph neural networks and propose a general \textbf{graph multi-task learning} framework in which different tasks can communicate with each other in an effective and interpretable way. We conduct extensive experiments in text classification and sequence labeling to evaluate our approach on multi-task learning and transfer learning. The empirical results show that our models not only outperform competitive baselines but also learn interpretable and transferable patterns across tasks.

cs.CL↗

arXiv2018

Contextualized Non-local Neural Networks for Sequence Learning

Recently, a large number of neural mechanisms and models have been proposed for sequence learning, of which self-attention, as exemplified by the Transformer model, and graph neural networks (GNNs) have attracted much attention. In this paper, we propose an approach that combines and draws on the complementary strengths of these two methods. Specifically, we propose contextualized non-local neural networks (CN$^{\textbf{3}}$), which can both dynamically construct a task-specific structure of a sentence and leverage rich local dependencies within a particular neighborhood. Experimental results on ten NLP tasks in text classification, semantic matching, and sequence labeling show that our proposed model outperforms competitive baselines and discovers task-specific dependency structures, thus providing better interpretability to users.

cs.CL↗

arXiv2018

Meta-Learning Multi-task Communication

In this paper, we describe a general framework: Parameters Read-Write Networks (PRaWNs) to systematically analyze current neural models for multi-task learning, in which we find that existing models expect to disentangle features into different spaces while features learned in practice are still entangled in shared space, leaving potential hazards for other training or unseen tasks. We propose to alleviate this problem by incorporating an inductive bias into the process of multi-task learning, that each task can keep informed of not only the knowledge stored in other tasks but the way how other tasks maintain their knowledge. In practice, we achieve above inductive bias by allowing different tasks to communicate by passing both hidden variables and gradients explicitly. Experimentally, we evaluate proposed methods on three groups of tasks and two types of settings (\textsc{in-task} and \textsc{out-of-task}). Quantitative and qualitative results show their effectiveness.

cs.CL↗

arXiv2018

Exploiting Effective Representations for Chinese Sentiment Analysis Using a Multi-Channel Convolutional Neural Network

Effective representation of a text is critical for various natural language processing tasks. For the particular task of Chinese sentiment analysis, it is important to understand and choose an effective representation of a text from different forms of Chinese representations such as word, character and pinyin. This paper presents a systematic study of the effect of these representations for Chinese sentiment analysis by proposing a multi-channel convolutional neural network (MCCNN), where each channel corresponds to a representation. Experimental results show that: (1) Word wins on the dataset of low OOV rate while character wins otherwise; (2) Using these representations in combination generally improves the performance; (3) The representations based on MCCNN outperform conventional ngram features using SVM; (4) The proposed MCCNN model achieves the competitive performance against the state-of-the-art model fastText for Chinese sentiment analysis.

cs.CL↗

arXiv2018

Meta Multi-Task Learning for Sequence Modeling

Semantic composition functions have been playing a pivotal role in neural representation learning of text sequences. In spite of their success, most existing models suffer from the underfitting problem: they use the same shared compositional function on all the positions in the sequence, thereby lacking expressive power due to incapacity to capture the richness of compositionality. Besides, the composition functions of different tasks are independent and learned from scratch. In this paper, we propose a new sharing scheme of composition function across multiple tasks. Specifically, we use a shared meta-network to capture the meta-knowledge of semantic composition and generate the parameters of the task-specific semantic composition models. We conduct extensive experiments on two types of tasks, text classification and sequence tagging, which demonstrate the benefits of our approach. Besides, we show that the shared meta-knowledge learned by our proposed model can be regarded as off-the-shelf knowledge and easily transferred to new tasks.

cs.AI↗