arXiv subjects

Jin Xu

Publications and source records attributed to Jin Xu.

At least 163 records · Page 9Linked to original sources

arXiv2022

Procedural Text Understanding via Scene-Wise Evolution

Procedural text understanding requires machines to reason about entity states within the dynamical narratives. Current procedural text understanding approaches are commonly \textbf{entity-wise}, which separately track each entity and independently predict different states of each entity. Such an entity-wise paradigm does not consider the interaction between entities and their states. In this paper, we propose a new \textbf{scene-wise} paradigm for procedural text understanding, which jointly tracks states of all entities in a scene-by-scene manner. Based on this paradigm, we propose \textbf{S}cene \textbf{G}raph \textbf{R}easoner (\textbf{SGR}), which introduces a series of dynamically evolving scene graphs to jointly formulate the evolution of entities, states and their associations throughout the narrative. In this way, the deep interactions between all entities and states can be jointly captured and simultaneously derived from scene graphs. Experiments show that SGR not only achieves the new state-of-the-art performance but also significantly accelerates the speed of reasoning.

cs.CL↗

arXiv2022

libRoadRunner 2.0: A High-Performance SBML Simulation and Analysis Library

Motivation: This paper presents libRoadRunner 2.0, an extensible, high-performance, cross-platform, open-source software library for the simulation and analysis of models expressed using Systems Biology Markup Language SBML). Results: libRoadRunner is a self-contained library, able to run both as a component inside other tools via its C++ and C bindings, and interactively through its Python or Julia interface. libRoadRunner uses a custom Just-In-Time JIT compiler built on the widely-used LLVM JIT compiler framework. It compiles SBML-specified models directly into native machine code for a large variety of processors, making it appropriate for solving extremely large models or repeated runs. libRoadRunner is flexible, supporting the bulk of the SBML specification (except for delay and nonlinear algebraic equations) and including several SBML extensions such as composition and distributions. It offers multiple deterministic and stochastic integrators, as well as tools for steady-state, sensitivity, stability analysis, and structural analysis of the stoichiometric matrix. Availability: libRoadRunner binary distributions are available for Mac OS X, Linux, and Windows. The library is licensed under the Apache License Version 2.0. libRoadRunner is also available for ARM-based computers such as the Raspberry Pi and can in principle be compiled on any system supported by LLVM-13. http://sys-bio.github.io/roadrunner/index.html provides online documentation, full build instructions, binaries, and a git source repository.

q-bio.QM↗

arXiv2022

CLS: Cross Labeling Supervision for Semi-Supervised Learning

It is well known that the success of deep neural networks is greatly attributed to large-scale labeled datasets. However, it can be extremely time-consuming and laborious to collect sufficient high-quality labeled data in most practical applications. Semi-supervised learning (SSL) provides an effective solution to reduce the cost of labeling by simultaneously leveraging both labeled and unlabeled data. In this work, we present Cross Labeling Supervision (CLS), a framework that generalizes the typical pseudo-labeling process. Based on FixMatch, where a pseudo label is generated from a weakly-augmented sample to teach the prediction on a strong augmentation of the same input sample, CLS allows the creation of both pseudo and complementary labels to support both positive and negative learning. To mitigate the confirmation bias of self-labeling and boost the tolerance to false labels, two different initialized networks with the same structure are trained simultaneously. Each network utilizes high-confidence labels from the other network as additional supervision signals. During the label generation phase, adaptive sample weights are assigned to artificial labels according to their prediction confidence. The sample weight plays two roles: quantify the generated labels' quality and reduce the disruption of inaccurate labels on network training. Experimental results on the semi-supervised classification task show that our framework outperforms existing approaches by large margins on the CIFAR-10 and CIFAR-100 datasets.

cs.CV↗

arXiv2022

Hybrid Contrastive Quantization for Efficient Cross-View Video Retrieval

With the recent boom of video-based social platforms (e.g., YouTube and TikTok), video retrieval using sentence queries has become an important demand and attracts increasing research attention. Despite the decent performance, existing text-video retrieval models in vision and language communities are impractical for large-scale Web search because they adopt brute-force search based on high-dimensional embeddings. To improve efficiency, Web search engines widely apply vector compression libraries (e.g., FAISS) to post-process the learned embeddings. Unfortunately, separate compression from feature encoding degrades the robustness of representations and incurs performance decay. To pursue a better balance between performance and efficiency, we propose the first quantized representation learning method for cross-view video retrieval, namely Hybrid Contrastive Quantization (HCQ). Specifically, HCQ learns both coarse-grained and fine-grained quantizations with transformers, which provide complementary understandings for texts and videos and preserve comprehensive semantic information. By performing Asymmetric-Quantized Contrastive Learning (AQ-CL) across views, HCQ aligns texts and videos at coarse-grained and multiple fine-grained levels. This hybrid-grained learning strategy serves as strong supervision on the cross-view video quantization model, where contrastive learning at different levels can be mutually promoted. Extensive experiments on three Web video benchmark datasets demonstrate that HCQ achieves competitive performance with state-of-the-art non-compressed retrieval methods while showing high efficiency in storage and computation. Code and configurations are available at https://github.com/gimpong/WWW22-HCQ.

cs.IR↗

arXiv2021

Age of Information for Single Buffer Systems with Vacation Server

In this research, we study the information freshness in M/G/1 queueing system with a single buffer and the server taking multiple vacations. This system has wide applications in communication systems. We aim to evaluate the information freshness in this system with both i.i.d. and non-i.i.d. vacations under three different scheduling policies, namely Conventional Buffer System (CBS), Buffer Relaxation System (BRS), and Conventional Buffer System with Preemption in Service (CBS-P). For the systems with i.i.d. vacations, we derive the closed-form expressions of information freshness metrics such as the expected Age of Information (AoI), the expected Peak Age of Information (PAoI), and the variance of peak age under each policy. For systems with non-i.i.d. vacations, we use the polling system as an example and provide the closed-form expression of its PAoI under each policy. We explore the conditions under which one of these policies has advantages over the others for each information freshness metric. We further perform numerical studies to validate our results and develop insights.

cs.PF↗

arXiv2021

AutoHEnsGNN: Winning Solution to AutoGraph Challenge for KDD Cup 2020

Graph Neural Networks (GNNs) have become increasingly popular and achieved impressive results in many graph-based applications. However, extensive manual work and domain knowledge are required to design effective architectures, and the results of GNN models have high variance with different training setups, which limits the application of existing GNN models. In this paper, we present AutoHEnsGNN, a framework to build effective and robust models for graph tasks without any human intervention. AutoHEnsGNN won first place in the AutoGraph Challenge for KDD Cup 2020, and achieved the best rank score of five real-life datasets in the final phase. Given a task, AutoHEnsGNN first applies a fast proxy evaluation to automatically select a pool of promising GNN models. Then it builds a hierarchical ensemble framework: 1) We propose graph self-ensemble (GSE), which can reduce the variance of weight initialization and efficiently exploit the information of local and global neighborhoods; 2) Based on GSE, a weighted ensemble of different types of GNN models is used to effectively learn more discriminative node representations. To efficiently search the architectures and ensemble weights, we propose AutoHEnsGNN$_{\text{Gradient}}$, which treats the architectures and ensemble weights as architecture parameters and uses gradient-based architecture search to obtain optimal configurations, and AutoHEnsGNN$_{\text{Adaptive}}$, which can adaptively adjust the ensemble weight based on the model accuracy. Extensive experiments on node classification, graph classification, edge prediction and KDD Cup challenge demonstrate the effectiveness and generality of AutoHEnsGNN

cs.LG↗

arXiv2021

55- and 56-configurations are reducible

Let $G$ be a 4-chromatic maximal planar graph (MPG) with the minimum degree of at least 4 and let $C$ be an even-length cycle of $G$.If $|f(C)|=2$ for every $f$ in some Kempe equivalence class of $G$, then we call $C$ an unchanged bichromatic cycle (UBC) of $G$, and correspondingly $G$ an unchanged bichromatic cycle maximal planar graph (UBCMPG) with respect to $C$, where $f(C)=\{f(v)| v\in V(C)\}$. For an UBCMPG $G$ with respect to an UBC $C$, the subgraph of $G$ induced by the set of edges belonging to $C$ and its interior (or exterior), denoted by $G^C$, is called a base-module of $G$; in particular, when the length of $C$ is equal to four, we use $C_4$ instead of $C$ and call $G^{C_4}$ a 4-base-module. In this paper, we first study the properties of UBCMPGs and show that every 4-base-module $G^{C_4}$ contains a 4-coloring under which $C_4$ is bichromatic and there are at least two bichromatic paths with different colors between one pair of diagonal vertices of $C_4$ (these paths are called module-paths). We further prove that every 4-base-module $G^{C_4}$ contains a 4-coloring (called decycle coloring) for which the ends of a module-path are colored by distinct colors. Finally, based on the technique of the contracting and extending operations of MPGs, we prove that 55-configurations and 56-configurations are reducible by converting the reducibility problem of these two classes of configurations into the decycle coloring problem of 4-base-modules.

math.CO↗

arXiv2021

Knowledgeable or Educated Guess? Revisiting Language Models as Knowledge Bases

Previous literatures show that pre-trained masked language models (MLMs) such as BERT can achieve competitive factual knowledge extraction performance on some datasets, indicating that MLMs can potentially be a reliable knowledge source. In this paper, we conduct a rigorous study to explore the underlying predicting mechanisms of MLMs over different extraction paradigms. By investigating the behaviors of MLMs, we find that previous decent performance mainly owes to the biased prompts which overfit dataset artifacts. Furthermore, incorporating illustrative cases and external contexts improve knowledge prediction mainly due to entity type guidance and golden answer leakage. Our findings shed light on the underlying predicting mechanisms of MLMs, and strongly question the previous conclusion that current MLMs can potentially serve as reliable factual knowledge bases.

cs.CL↗

arXiv2021

Text2Event: Controllable Sequence-to-Structure Generation for End-to-end Event Extraction

Event extraction is challenging due to the complex structure of event records and the semantic gap between text and event. Traditional methods usually extract event records by decomposing the complex structure prediction task into multiple subtasks. In this paper, we propose Text2Event, a sequence-to-structure generation paradigm that can directly extract events from the text in an end-to-end manner. Specifically, we design a sequence-to-structure network for unified event extraction, a constrained decoding algorithm for event knowledge injection during inference, and a curriculum learning algorithm for efficient model learning. Experimental results show that, by uniformly modeling all tasks in a single model and universally predicting different labels, our method can achieve competitive performance using only record-level annotations in both supervised learning and transfer learning settings.

cs.CL↗

arXiv2021

From Discourse to Narrative: Knowledge Projection for Event Relation Extraction

Current event-centric knowledge graphs highly rely on explicit connectives to mine relations between events. Unfortunately, due to the sparsity of connectives, these methods severely undermine the coverage of EventKGs. The lack of high-quality labelled corpora further exacerbates that problem. In this paper, we propose a knowledge projection paradigm for event relation extraction: projecting discourse knowledge to narratives by exploiting the commonalities between them. Specifically, we propose Multi-tier Knowledge Projection Network (MKPNet), which can leverage multi-tier discourse knowledge effectively for event relation extraction. In this way, the labelled data requirement is significantly reduced, and implicit event relations can be effectively extracted. Intrinsic experimental results show that MKPNet achieves the new state-of-the-art performance, and extrinsic experimental results verify the value of the extracted event relations.

cs.CL↗

arXiv2021

Group Equivariant Subsampling

Subsampling is used in convolutional neural networks (CNNs) in the form of pooling or strided convolutions, to reduce the spatial dimensions of feature maps and to allow the receptive fields to grow exponentially with depth. However, it is known that such subsampling operations are not translation equivariant, unlike convolutions that are translation equivariant. Here, we first introduce translation equivariant subsampling/upsampling layers that can be used to construct exact translation equivariant CNNs. We then generalise these layers beyond translations to general groups, thus proposing group equivariant subsampling/upsampling. We use these layers to construct group equivariant autoencoders (GAEs) that allow us to learn low-dimensional equivariant representations. We empirically verify on images that the representations are indeed equivariant to input translations and rotations, and thus generalise well to unseen positions and orientations. We further use GAEs in models that learn object-centric representations on multi-object datasets, and show improved data efficiency and decomposition compared to non-equivariant baselines.

cs.LG↗

arXiv2021

NAS-BERT: Task-Agnostic and Adaptive-Size BERT Compression with Neural Architecture Search

While pre-trained language models (e.g., BERT) have achieved impressive results on different natural language processing tasks, they have large numbers of parameters and suffer from big computational and memory costs, which make them difficult for real-world deployment. Therefore, model compression is necessary to reduce the computation and memory cost of pre-trained models. In this work, we aim to compress BERT and address the following two challenging practical issues: (1) The compression algorithm should be able to output multiple compressed models with different sizes and latencies, in order to support devices with different memory and latency limitations; (2) The algorithm should be downstream task agnostic, so that the compressed models are generally applicable for different downstream tasks. We leverage techniques in neural architecture search (NAS) and propose NAS-BERT, an efficient method for BERT compression. NAS-BERT trains a big supernet on a search space containing a variety of architectures and outputs multiple compressed models with adaptive sizes and latency. Furthermore, the training of NAS-BERT is conducted on standard self-supervised pre-training tasks (e.g., masked language model) and does not depend on specific downstream tasks. Thus, the compressed models can be used across various downstream tasks. The technical challenge of NAS-BERT is that training a big supernet on the pre-training task is extremely costly. We employ several techniques including block-wise search, search space pruning, and performance approximation to improve search efficiency and accuracy. Extensive experiments on GLUE and SQuAD benchmark datasets demonstrate that NAS-BERT can find lightweight models with better accuracy than previous approaches, and can be directly applied to different downstream tasks with adaptive model sizes for different requirements of memory or latency.

cs.CL↗

arXiv2021

The Graph-Based Behavior-Aware Recommendation for Interactive News

Interactive news recommendation has been launched and attracted much attention recently. In this scenario, user's behavior evolves from single click behavior to multiple behaviors including like, comment, share etc. However, most of the existing methods still use single click behavior as the unique criterion of judging user's preferences. Further, although heterogeneous graphs have been applied in different areas, a proper way to construct a heterogeneous graph for interactive news data with an appropriate learning mechanism on it is still desired. To address the above concerns, we propose a graph-based behavior-aware network, which simultaneously considers six different types of behaviors as well as user's demand on the news diversity. We have three main steps. First, we build an interaction behavior graph for multi-level and multi-category data. Second, we apply DeepWalk on the behavior graph to obtain entity semantics, then build a graph-based convolutional neural network called G-CNN to learn news representations, and an attention-based LSTM to learn behavior sequence representations. Third, we introduce core and coritivity features for the behavior graph, which measure the concentration degree of user's interests. These features affect the trade-off between accuracy and diversity of our personalized recommendation system. Taking these features into account, our system finally achieves recommending news to different users at their different levels of concentration degrees.

cs.IR↗

arXiv2021

MixSpeech: Data Augmentation for Low-resource Automatic Speech Recognition

In this paper, we propose MixSpeech, a simple yet effective data augmentation method based on mixup for automatic speech recognition (ASR). MixSpeech trains an ASR model by taking a weighted combination of two different speech features (e.g., mel-spectrograms or MFCC) as the input, and recognizing both text sequences, where the two recognition losses use the same combination weight. We apply MixSpeech on two popular end-to-end speech recognition models including LAS (Listen, Attend and Spell) and Transformer, and conduct experiments on several low-resource datasets including TIMIT, WSJ, and HKUST. Experimental results show that MixSpeech achieves better accuracy than the baseline models without data augmentation, and outperforms a strong data augmentation method SpecAugment on these recognition tasks. Specifically, MixSpeech outperforms SpecAugment with a relative PER improvement of 10.6$\%$ on TIMIT dataset, and achieves a strong WER of 4.7$\%$ on WSJ dataset.

cs.CL↗

arXiv2021

Studentized Permutation Method for Comparing Restricted Mean Survival Times with Small Sample from Randomized Trials

Recent observations, especially in cancer immunotherapy clinical trials with time-to-event outcomes, show that the commonly used proportial hazard assumption is often not justifiable, hampering an appropriate analyse of the data by hazard ratios. An attractive alternative advocated is given by the restricted mean survival time (RMST), which does not rely on any model assumption and can always be interpreted intuitively. As pointed out recently by Horiguchi and Uno (2020), methods for the RMST based on asymptotic theory suffer from inflated type-I error under small sample sizes. To overcome this problem, they suggested a permutation strategy leading to more convincing results in simulations. However, their proposal requires an exchangeable data set-up between comparison groups which may be limiting in practice. In addition, it is not possible to invert their testing procedure to obtain valid confidence intervals, which can provide more in-depth information. In this paper, we address these limitations by proposing a studentized permutation test as well as the corresponding permutation-based confidence intervals. In our extensive simulation study, we demonstrate the advantage of our new method, especially in situations with relative small sample sizes and unbalanced groups. Finally we illustrate the application of the proposed method by re-analysing data from a recent lung cancer clinical trial.

stat.ME↗

arXiv2020

Broadband directional control of thermal emission

Controlling the directionality of emitted far-field thermal radiation is a fundamental challenge in contemporary photonics and materials research. While photonic strategies have enabled angular selectivity of thermal emission over narrow sets of bandwidths, thermal radiation is inherently a broadband phenomenon. We currently lack the ability to constrain emitted thermal radiation to arbitrary angular ranges over broad bandwidths. Here, we introduce and experimentally realize gradient epsilon-near-zero (ENZ) material structures that enable broad spectrum directional control of thermal emission by supporting leaky electromagnetic modes that couple to free space at fixed angles over a broad bandwidth. We demonstrate two emitter structures consisting of multiple semiconductor oxides in a photonic configuration that enable gradient ENZ behavior over long-wave infrared wavelengths. The structures exhibit high average emissivity (greater than 0.6 and 0.7) in the p polarization between 7.7 and 11.5 micron over an angular range of 70 deg - 85 deg, and between 10.0 to 14.3 micron over an angular range of 60 deg-75 deg, respectively. Outside these angular ranges, the emissivity dramatically drops to 0.4 at 50 deg and 40 deg. The structures broadband thermal beaming capability enables strong radiative heat transfer only at particular angles and is experimentally verified through direct measurements of thermal emission. By decoupling conventional limitations on angular and spectral response, our approach opens new possibilities for radiative heat transfer in applications such as thermal camouflaging, solar heating, radiative cooling and waste heat recovery.

physics.optics↗

arXiv2020

Peak Age of Information in Priority Queueing Systems

We consider a priority queueing system where a single processor serves k classes of packets that are generated randomly following Poisson processes. Our objective is to compute the expected Peak Age of Information (PAoI) under various scenarios. In particular, we consider two situations where the buffer size at each queue is one and infinite, and in the infinite buffer size case we consider First Come First Serve (FCFS) and Last Come First Serve (LCFS) as service disciplines. For the system with buffer size one at each queue, we derive PAoI exactly for the case of exponential service time and bounds (which are excellent approximations) for the case of general service time, with small k. For the system with infinite buffer size, we provide closed-form expressions of PAoI for both FCFS and LCFS where service time is general and k could be large. Using those results we investigated the effect of ordering of priorities and service disciplines for the various scenarios. We perform extensive numerical studies to validate our results and develop insights.

cs.IT↗

arXiv2020

Incorporate Semantic Structures into Machine Translation Evaluation via UCCA

Copying mechanism has been commonly used in neural paraphrasing networks and other text generation tasks, in which some important words in the input sequence are preserved in the output sequence. Similarly, in machine translation, we notice that there are certain words or phrases appearing in all good translations of one source text, and these words tend to convey important semantic information. Therefore, in this work, we define words carrying important semantic meanings in sentences as semantic core words. Moreover, we propose an MT evaluation approach named Semantically Weighted Sentence Similarity (SWSS). It leverages the power of UCCA to identify semantic core words, and then calculates sentence similarity scores on the overlap of semantic core words. Experimental results show that SWSS can consistently improve the performance of popular MT evaluation metrics which are based on lexical similarity.

cs.CL↗