arXiv subjects

Xu Han

Publications and source records attributed to Xu Han.

At least 289 records · Page 16Linked to original sources

arXiv2021

Large-Scale Multiple Testing for Matrix-Valued Data under Double Dependency

High-dimensional inference based on matrix-valued data has drawn increasing attention in modern statistical research, yet not much progress has been made in large-scale multiple testing specifically designed for analysing such data sets. Motivated by this, we consider in this article an electroencephalography (EEG) experiment that produces matrix-valued data and presents a scope of developing novel matrix-valued data based multiple testing methods controlling false discoveries for hypotheses that are of importance in such an experiment. The row-column cross-dependency of observations appearing in a matrix form, referred to as double-dependency, is one of the main challenges in the development of such methods. We address it by assuming matrix normal distribution for the observations at each of the independent matrix data-points. This allows us to fully capture the underlying double-dependency informed through the row- and column-covariance matrices and develop methods that are potentially more powerful than the corresponding one (e.g., Fan and Han (2017)) obtained by vectorizing each data point and thus ignoring the double-dependency. We propose two methods to approximate the false discovery proportion with statistical accuracy. While one of these methods is a general approach under double-dependency, the other one provides more computational efficiency for higher dimensionality. Extensive numerical studies illustrate the superior performance of the proposed methods over the principal factor approximation method of Fan and Han (2017). The proposed methods have been further applied to the aforementioned EEG data.

stat.ME↗

arXiv2021

Pre-processing with Orthogonal Decompositions for High-dimensional Explanatory Variables

Strong correlations between explanatory variables are problematic for high-dimensional regularized regression methods. Due to the violation of the Irrepresentable Condition, the popular LASSO method may suffer from false inclusions of inactive variables. In this paper, we propose pre-processing with orthogonal decompositions (PROD) for the explanatory variables in high-dimensional regressions. The PROD procedure is constructed based upon a generic orthogonal decomposition of the design matrix. We demonstrate by two concrete cases that the PROD approach can be effectively constructed for improving the performance of high-dimensional penalized regression. Our theoretical analysis reveals their properties and benefits for high-dimensional penalized linear regression with LASSO. Extensive numerical studies with simulations and data analysis show the promising performance of the PROD.

stat.ME↗

arXiv2021

Order Matters: Probabilistic Modeling of Node Sequence for Graph Generation

A graph generative model defines a distribution over graphs. One type of generative model is constructed by autoregressive neural networks, which sequentially add nodes and edges to generate a graph. However, the likelihood of a graph under the autoregressive model is intractable, as there are numerous sequences leading to the given graph; this makes maximum likelihood estimation challenging. Instead, in this work we derive the exact joint probability over the graph and the node ordering of the sequential process. From the joint, we approximately marginalize out the node orderings and compute a lower bound on the log-likelihood using variational inference. We train graph generative models by maximizing this bound, without using the ad-hoc node orderings of previous methods. Our experiments show that the log-likelihood bound is significantly tighter than the bound of previous schemes. Moreover, the models fitted with the proposed algorithm can generate high-quality graphs that match the structures of target graphs not seen during training. We have made our code publicly available at \hyperref[https://github.com/tufts-ml/graph-generation-vi]{https://github.com/tufts-ml/graph-generation-vi}.

stat.ML↗

arXiv2021

Mutual Information Regularized Identity-aware Facial ExpressionRecognition in Compressed Video

How to extract effective expression representations that invariant to the identity-specific attributes is a long-lasting problem for facial expression recognition (FER). Most of the previous methods process the RGB images of a sequence, while we argue that the off-the-shelf and valuable expression-related muscle movement is already embedded in the compression format. In this paper, we target to explore the inter-subject variations eliminated facial expression representation in the compressed video domain. In the up to two orders of magnitude compressed domain, we can explicitly infer the expression from the residual frames and possibly extract identity factors from the I frame with a pre-trained face recognition network. By enforcing the marginal independence of them, the expression feature is expected to be purer for the expression and be robust to identity shifts. Specifically, we propose a novel collaborative min-min game for mutual information (MI) minimization in latent space. We do not need the identity label or multiple expression samples from the same person for identity elimination. Moreover, when the apex frame is annotated in the dataset, the complementary constraint can be further added to regularize the feature-level game. In testing, only the compressed residual frames are required to achieve expression prediction. Our solution can achieve comparable or better performance than the recent decoded image-based methods on the typical FER benchmarks with about 3 times faster inference.

cs.CV↗

arXiv2021

CLEVE: Contrastive Pre-training for Event Extraction

Event extraction (EE) has considerably benefited from pre-trained language models (PLMs) by fine-tuning. However, existing pre-training methods have not involved modeling event characteristics, resulting in the developed EE models cannot take full advantage of large-scale unsupervised data. To this end, we propose CLEVE, a contrastive pre-training framework for EE to better learn event knowledge from large unsupervised data and their semantic structures (e.g. AMR) obtained with automatic parsers. CLEVE contains a text encoder to learn event semantics and a graph encoder to learn event structures respectively. Specifically, the text encoder learns event semantic representations by self-supervised contrastive learning to represent the words of the same events closer than those unrelated words; the graph encoder learns event structure representations by graph contrastive pre-training on parsed event-related semantic structures. The two complementary representations then work together to improve both the conventional supervised EE and the unsupervised "liberal" EE, which requires jointly extracting events and discovering event schemata without any annotated data. Experiments on ACE 2005 and MAVEN datasets show that CLEVE achieves significant improvements, especially in the challenging unsupervised setting. The source code and pre-trained checkpoints can be obtained from https://github.com/THU-KEG/CLEVE.

cs.CL↗

arXiv2021

Manual Evaluation Matters: Reviewing Test Protocols of Distantly Supervised Relation Extraction

Distantly supervised (DS) relation extraction (RE) has attracted much attention in the past few years as it can utilize large-scale auto-labeled data. However, its evaluation has long been a problem: previous works either took costly and inconsistent methods to manually examine a small sample of model predictions, or directly test models on auto-labeled data -- which, by our check, produce as much as 53% wrong labels at the entity pair level in the popular NYT10 dataset. This problem has not only led to inaccurate evaluation, but also made it hard to understand where we are and what's left to improve in the research of DS-RE. To evaluate DS-RE models in a more credible way, we build manually-annotated test sets for two DS-RE datasets, NYT10 and Wiki20, and thoroughly evaluate several competitive models, especially the latest pre-trained ones. The experimental results show that the manual evaluation can indicate very different conclusions from automatic ones, especially some unexpected observations, e.g., pre-trained models can achieve dominating performance while being more susceptible to false-positives compared to previous methods. We hope that both our manual test sets and novel observations can help advance future DS-RE research.

cs.CL↗

arXiv2021

Jump on the Bandwagon? -- Characterizing Bandwagon Phenomenon in Online NBA Fan Communities

Understanding user dynamics in online communities has become an active research topic and can provide valuable insights for human behavior analysis and community management. In this work, we investigate the "bandwagon fan" phenomenon, a special case of user dynamics, to provide a large-scale characterization of online fan loyalty in the context of professional sports teams. We leverage the existing structure of NBA-related discussion forums on Reddit, investigate the general bandwagon patterns, and trace the behavior of bandwagon fans to capture latent behavioral characteristics. We observe that better teams attract more bandwagon fans, but they do not necessarily come from weak teams. Our analysis of bandwagon fan flow also shows different trends for different teams, as the playoff season progresses. Furthermore, we compare bandwagon users with non-bandwagon users in terms of their activity and language usage. We find that bandwagon users write shorter comments but receive better feedback, and use words that show less attachment to their affiliated teams. Our observations allow for more effective identification of bandwagon users and prediction of users' future bandwagon behavior in a season, as demonstrated by the significant improvement over the baseline method in our evaluation results.

cs.CY↗

arXiv2021

Designing Effective Interview Chatbots: Automatic Chatbot Profiling and Design Suggestion Generation for Chatbot Debugging

Recent studies show the effectiveness of interview chatbots for information elicitation. However, designing an effective interview chatbot is non-trivial. Few tools exist to help designers design, evaluate, and improve an interview chatbot iteratively. Based on a formative study and literature reviews, we propose a computational framework for quantifying the performance of interview chatbots. Incorporating the framework, we have developed iChatProfile, an assistive chatbot design tool that can automatically generate a profile of an interview chatbot with quantified performance metrics and offer design suggestions for improving the chatbot based on such metrics. To validate the effectiveness of iChatProfile, we designed and conducted a between-subject study that compared the performance of 10 interview chatbots designed with or without using iChatProfile. Based on the live chats between the 10 chatbots and 1349 users, our results show that iChatProfile helped the designers build significantly more effective interview chatbots, improving both interview quality and user experience.

cs.HC↗

arXiv2021

Quasi-maximum likelihood estimation of break point in high-dimensional factor models

This paper estimates the break point for large-dimensional factor models with a single structural break in factor loadings at a common unknown date. First, we propose a quasi-maximum likelihood (QML) estimator of the change point based on the second moments of factors, which are estimated by principal component analysis. We show that the QML estimator performs consistently when the covariance matrix of the pre- or post-break factor loading, or both, is singular. When the loading matrix undergoes a rotational type of change while the number of factors remains constant over time, the QML estimator incurs a stochastically bounded estimation error. In this case, we establish an asymptotic distribution of the QML estimator. The simulation results validate the feasibility of this estimator when used in finite samples. In addition, we demonstrate empirical applications of the proposed method by applying it to estimate the break points in a U.S. macroeconomic dataset and a stock return dataset.

econ.EM↗

arXiv2021

Automatic Generation of Large-scale 3D Road Networks based on GIS Data

How to automatically generate a realistic large-scale 3D road network is a key point for immersive and credible traffic simulations. Existing methods cannot automatically generate various kinds of intersections in 3D space based on GIS data. In this paper, we propose a method to generate complex and large-scale 3D road networks automatically with the open source GIS data, including satellite imagery, elevation data and two-dimensional(2D) road center axis data, as input. We first introduce a semantic structure of road network to obtain high-detailed and well-formed networks in a 3D scene. We then generate 2D shapes and topological data of the road network according to the semantic structure and 2D road center axis data. At last, we segment the elevation data and generate the surface of the 3D road network according to the 2D semantic data and satellite imagery data. Results show that our method does well in the generation of various types of intersections and the high-detailed features of roads. The traffic semantic structure, which must be provided in traffic simulation, can also be generated automatically according to our method.

cs.GR↗

arXiv2021

CSS-LM: A Contrastive Framework for Semi-supervised Fine-tuning of Pre-trained Language Models

Fine-tuning pre-trained language models (PLMs) has demonstrated its effectiveness on various downstream NLP tasks recently. However, in many low-resource scenarios, the conventional fine-tuning strategies cannot sufficiently capture the important semantic features for downstream tasks. To address this issue, we introduce a novel framework (named "CSS-LM") to improve the fine-tuning phase of PLMs via contrastive semi-supervised learning. Specifically, given a specific task, we retrieve positive and negative instances from large-scale unlabeled corpora according to their domain-level and class-level semantic relatedness to the task. We then perform contrastive semi-supervised learning on both the retrieved unlabeled and original labeled instances to help PLMs capture crucial task-related semantic features. The experimental results show that CSS-LM achieves better results than the conventional fine-tuning strategy on a series of downstream tasks with few-shot settings, and outperforms the latest supervised contrastive fine-tuning strategies. Our datasets and source code will be available to provide more details.

cs.CL↗

arXiv2021

Identity-aware Facial Expression Recognition in Compressed Video

This paper targets to explore the inter-subject variations eliminated facial expression representation in the compressed video domain. Most of the previous methods process the RGB images of a sequence, while the off-the-shelf and valuable expression-related muscle movement already embedded in the compression format. In the up to two orders of magnitude compressed domain, we can explicitly infer the expression from the residual frames and possible to extract identity factors from the I frame with a pre-trained face recognition network. By enforcing the marginal independent of them, the expression feature is expected to be purer for the expression and be robust to identity shifts. We do not need the identity label or multiple expression samples from the same person for identity elimination. Moreover, when the apex frame is annotated in the dataset, the complementary constraint can be further added to regularize the feature-level game. In testing, only the compressed residual frames are required to achieve expression prediction. Our solution can achieve comparable or better performance than the recent decoded image based methods on the typical FER benchmarks with about 3$\times$ faster inference with compressed data.

cs.CV↗

arXiv2021

Inverse Faraday Effect in an Optomagnonic Waveguide

Single-mode high-index-contrast waveguides have been ubiquitously exploited in optical, microwave, and phononic structures for achieving enhanced wave-matter interactions. Although micro-scale optomechanical and electro-optical devices have been widely studied, optomagnonic devices remain a grand challenge at the microscale. Here, we introduce a planar optomagnonic waveguide platform based on a ferrimagnetic insulator that simultaneously supports single transverse mode of spin waves (magnons) and highly confined optical modes. The co-localization of spin and light waves gives rise to enhanced inverse Faraday effect, and as a result, magnons are excited by an effective magnetic field generated by interacting optical photons. Moreover, the strongly enhanced optomagnonic interaction allows us to observe such effect using low-power (milliwatt level) light signals in the continuous-wave form, as opposed to high-intensity (megawatt peak power) light pulses that are typically required in magnetic bulk materials or thin films. The optically-driven magnons are detected electrically with preserved phase coherence, showing the feasibility for launching spin waves with low-power continuous optical fields.

physics.optics↗

arXiv2020

Adversarial Language Games for Advanced Natural Language Intelligence

We study the problem of adversarial language games, in which multiple agents with conflicting goals compete with each other via natural language interactions. While adversarial language games are ubiquitous in human activities, little attention has been devoted to this field in natural language processing. In this work, we propose a challenging adversarial language game called Adversarial Taboo as an example, in which an attacker and a defender compete around a target word. The attacker is tasked with inducing the defender to utter the target word invisible to the defender, while the defender is tasked with detecting the target word before being induced by the attacker. In Adversarial Taboo, a successful attacker must hide its intention and subtly induce the defender, while a competitive defender must be cautious with its utterances and infer the intention of the attacker. Such language abilities can facilitate many important downstream NLP tasks. To instantiate the game, we create a game environment and a competition platform. Comprehensive experiments and empirical studies on several baseline attack and defense strategies show promising and interesting results. Based on the analysis on the game and experiments, we discuss multiple promising directions for future research.

cs.CL↗

arXiv2020

GAN Ensemble for Anomaly Detection

When formulated as an unsupervised learning problem, anomaly detection often requires a model to learn the distribution of normal data. Previous works apply Generative Adversarial Networks (GANs) to anomaly detection tasks and show good performances from these models. Motivated by the observation that GAN ensembles often outperform single GANs in generation tasks, we propose to construct GAN ensembles for anomaly detection. In the proposed method, a group of generators and a group of discriminators are trained together, so every generator gets feedback from multiple discriminators, and vice versa. Compared to a single GAN, a GAN ensemble can better model the distribution of normal data and thus better detect anomalies. Our theoretical analysis of GANs and GAN ensembles explains the role of a GAN discriminator in anomaly detection. In the empirical study, we evaluate ensembles constructed from four types of base models, and the results show that these ensembles clearly outperform single models in a series of tasks of anomaly detection.

cs.LG↗

arXiv2020

Learning from Context or Names? An Empirical Study on Neural Relation Extraction

Neural models have achieved remarkable success on relation extraction (RE) benchmarks. However, there is no clear understanding which type of information affects existing RE models to make decisions and how to further improve the performance of these models. To this end, we empirically study the effect of two main information sources in text: textual context and entity mentions (names). We find that (i) while context is the main source to support the predictions, RE models also heavily rely on the information from entity mentions, most of which is type information, and (ii) existing datasets may leak shallow heuristics via entity mentions and thus contribute to the high performance on RE benchmarks. Based on the analyses, we propose an entity-masked contrastive pre-training framework for RE to gain a deeper understanding on both textual context and type information while avoiding rote memorization of entities or use of superficial cues in mentions. We carry out extensive experiments to support our views, and show that our framework can improve the effectiveness and robustness of neural models in different RE scenarios. All the code and datasets are released at https://github.com/thunlp/RE-Context-or-Names.

cs.CL↗

arXiv2020

CPM: A Large-scale Generative Chinese Pre-trained Language Model

Pre-trained Language Models (PLMs) have proven to be beneficial for various downstream NLP tasks. Recently, GPT-3, with 175 billion parameters and 570GB training data, drew a lot of attention due to the capacity of few-shot (even zero-shot) learning. However, applying GPT-3 to address Chinese NLP tasks is still challenging, as the training corpus of GPT-3 is primarily English, and the parameters are not publicly available. In this technical report, we release the Chinese Pre-trained Language Model (CPM) with generative pre-training on large-scale Chinese training data. To the best of our knowledge, CPM, with 2.6 billion parameters and 100GB Chinese training data, is the largest Chinese pre-trained language model, which could facilitate several downstream Chinese NLP tasks, such as conversation, essay generation, cloze test, and language understanding. Extensive experiments demonstrate that CPM achieves strong performance on many NLP tasks in the settings of few-shot (even zero-shot) learning. The code and parameters are available at https://github.com/TsinghuaAI/CPM-Generate.

cs.CL↗

arXiv2020

Deep Reinforcement Learning of Transition States

Combining reinforcement learning (RL) and molecular dynamics (MD) simulations, we propose a machine-learning approach (RL$^‡$) to automatically unravel chemical reaction mechanisms. In RL$^‡$, locating the transition state of a chemical reaction is formulated as a game, where a virtual player is trained to shoot simulation trajectories connecting the reactant and product. The player utilizes two functions, one for value estimation and the other for policy making, to iteratively improve the chance of winning this game. We can directly interpret the reaction mechanism according to the value function. Meanwhile, the policy function enables efficient sampling of the transition paths, which can be further used to analyze the reaction dynamics and kinetics. Through multiple experiments, we show that RL‡ can be trained tabula rasa hence allows us to reveal chemical reaction mechanisms with minimal subjective biases.

physics.chem-ph↗