arXiv subjects

Li Zhang

Publications and source records attributed to Li Zhang.

At least 721 records · Page 40Linked to original sources

arXiv2021

Boundary-sensitive Pre-training for Temporal Localization in Videos

Many video analysis tasks require temporal localization thus detection of content changes. However, most existing models developed for these tasks are pre-trained on general video action classification tasks. This is because large scale annotation of temporal boundaries in untrimmed videos is expensive. Therefore no suitable datasets exist for temporal boundary-sensitive pre-training. In this paper for the first time, we investigate model pre-training for temporal localization by introducing a novel boundary-sensitive pretext (BSP) task. Instead of relying on costly manual annotations of temporal boundaries, we propose to synthesize temporal boundaries in existing video action classification datasets. With the synthesized boundaries, BSP can be simply conducted via classifying the boundary types. This enables the learning of video representations that are much more transferable to downstream temporal localization tasks. Extensive experiments show that the proposed BSP is superior and complementary to the existing action classification based pre-training counterpart, and achieves new state-of-the-art performance on several temporal localization tasks.

cs.CV↗

arXiv2021

Robust and Accurate Object Detection via Adversarial Learning

Data augmentation has become a de facto component for training high-performance deep image classifiers, but its potential is under-explored for object detection. Noting that most state-of-the-art object detectors benefit from fine-tuning a pre-trained classifier, we first study how the classifiers' gains from various data augmentations transfer to object detection. The results are discouraging; the gains diminish after fine-tuning in terms of either accuracy or robustness. This work instead augments the fine-tuning stage for object detectors by exploring adversarial examples, which can be viewed as a model-dependent data augmentation. Our method dynamically selects the stronger adversarial images sourced from a detector's classification and localization branches and evolves with the detector to ensure the augmentation policy stays current and relevant. This model-dependent augmentation generalizes to different object detectors better than AutoAugment, a model-agnostic augmentation policy searched based on one particular detector. Our approach boosts the performance of state-of-the-art EfficientDets by +1.1 mAP on the COCO object detection benchmark. It also improves the detectors' robustness against natural distortions by +3.8 mAP and against domain shift by +1.3 mAP. Models are available at https://github.com/google/automl/tree/master/efficientdet/Det-AdvProp.md

cs.CV↗

arXiv2021

Learning Dynamic Alignment via Meta-filter for Few-shot Learning

Few-shot learning (FSL), which aims to recognise new classes by adapting the learned knowledge with extremely limited few-shot (support) examples, remains an important open problem in computer vision. Most of the existing methods for feature alignment in few-shot learning only consider image-level or spatial-level alignment while omitting the channel disparity. Our insight is that these methods would lead to poor adaptation with redundant matching, and leveraging channel-wise adjustment is the key to well adapting the learned knowledge to new classes. Therefore, in this paper, we propose to learn a dynamic alignment, which can effectively highlight both query regions and channels according to different local support information. Specifically, this is achieved by first dynamically sampling the neighbourhood of the feature position conditioned on the input few shot, based on which we further predict a both position-dependent and channel-dependent Dynamic Meta-filter. The filter is used to align the query feature with position-specific and channel-specific knowledge. Moreover, we adopt Neural Ordinary Differential Equation (ODE) to enable a more accurate control of the alignment. In such a sense our model is able to better capture fine-grained semantic context of the few-shot example and thus facilitates dynamical knowledge adaptation for few-shot learning. The resulting framework establishes the new state-of-the-arts on major few-shot visual recognition benchmarks, including miniImageNet and tieredImageNet.

cs.CV↗

arXiv2021

Fluence Adaptation for Task-based Dose Optimization in X-ray Phase-Contrast Imaging

Purpose: Grating-based imaging (GBI) and edge-illumination (EI) are two promising types of XPCI as the conventional x-ray sources can be directly utilized. For GBI and EI systems, the phase-stepping acquisition with multiple exposures at a constant fluence is usually adopted in the literature. This work, however, attempts to challenge such a constant fluence concept during the phase-stepping process and proposes a fluence adaptation mechanism for dose reduction. Method: Recently, analytic multi-order moment analysis has been proposed to improve the computing efficiency. In these algorithms, multiple contrasts can be calculated by summing together the weighted phase-stepping curves (PSCs) with some kernel functions, which suggests us that the raw data at different steps have different contributions for the noise in retrieved contrasts. Based on analytic retrieval formulas and the Gaussian noise model for detected signals, we derived an optimal adaptive fluence distribution, which is proportional to the absolute weighting kernel functions and the root of original sample PSCs acquired under the constant fluence. Results: To validate our analyses, simulations and experiments are conducted for GBI and EI systems. Simulated results demonstrate that the dose reduction ratio between our proposed fluence distributions and the typical constant one can be about 20% for the phase contrast, which is consistent with our theoretical predictions. Although the experimental noise reduction ratios are a little smaller than the theoretical ones, synthetic and real experiments both observe better noise performance by our proposed method. Our simulated results also give out the effective ranges of the parameters of the PSCs, such as the visibility in GBI, the standard deviation and the mean value in EI, providing a guidance for the use of our proposed approach in practice.

physics.med-ph↗

arXiv2021

Long-term multi-wavelength variations of Fermi blazar 3C 279

The long-term optical, X-ray and $γ$-ray data of blazar 3C 279 have been compiled from $Swift$-XRT, $RXTE$ PCA, $Fermi$-LAT, SMARTS and literature. The source exhibits strong variability on long time scales. Since 1980s to now, the optical $R$ band light curve spans above 32 yr, and a possible 5.6-yr-long quasi-periodic variation component has been found in it. The optical spectral behavior has been investigated. In the optical band, the mean spectral index is -1.71. The source exhibits an obvious special spectral behavior. In the low state, the source shows a clear bluer-when-brighter behavior in a sense that the optical spectrum turns harder (flatter) when the brightness increases. While in the high state, the optical spectrum is stable, that means the source spectral index does not vary with the brightness. The correlation analysis has been performed among optical, X-ray and $γ$-ray energy bands. The result indicates that the variations of $γ$-ray and X-ray bands are well correlated without time delay on the time scale of days, and their variations exhibit weak correlations with those of optical band. The variations, especial outbursts, are simultaneous, but the magnitude of variations is disproportionate. The detailed analysis reveals that the main outbursts exhibit strong correlations in different $γ$-ray, X-ray and optical bands.

astro-ph.HE↗

arXiv2021

Automatically detecting the conflicts between software requirements based on finer semantic analysis

Context: Conflicts between software requirements bring uncertainties to product development. Some great approaches have been proposed to identify these conflicts. However, they usually require the software requirements represented with specific templates and/or depend on other external source which is often uneasy to build for lots of projects in practice. Objective: We aim to propose an approach Finer Semantic Analysis-based Requirements Conflict Detector (FSARC) to automatically detecting the conflicts between the given natural language functional requirements by analyzing their finer semantic compositions. Method: We build a harmonized semantic meta-model of functional requirements with the form of eight-tuple. Then we propose algorithms to automatically analyze the linguistic features of requirements and to annotate the semantic elements for their semantic model construction. And we define seven types of conflicts as long as their heuristic detecting rules on the ground of their text pattern and semantical dependency. Finally, we design and implement the algorithm for conflicts detection. Results: The experiment with four requirement datasets illustrates that the recall of FSARC is nearly 100% and the average precision is 83.88% on conflicts detection. Conclusion: We provide a useful tool for detecting the conflicts between natural language functional requirements to improve the quality of the final requirements set. Besides, our approach is capable of transforming the natural language functional requirements into eight semantic tuples, which is useful not only the detection of the conflicts between requirements but also some other tasks such as constructing the association between requirements and so on.

cs.SE↗

arXiv2021

The NPU System for the 2020 Personalized Voice Trigger Challenge

This paper describes the system developed by the NPU team for the 2020 personalized voice trigger challenge. Our submitted system consists of two independently trained subsystems: a small footprint keyword spotting (KWS) system and a speaker verification (SV) system. For the KWS system, a multi-scale dilated temporal convolutional (MDTC) network is proposed to detect wake-up word (WuW). For SV system, Write something here. The KWS predicts posterior probabilities of whether an audio utterance contains WuW and estimates the location of WuW at the same time. When the posterior probability ofWuW reaches a predefined threshold, the identity information of triggered segment is determined by the SV system. On evaluation dataset, our submitted system obtains detection costs of 0.081and 0.091 in close talking and far-field tasks, respectively.

cs.SD↗

arXiv2021

Unlocking the Origin of Compositional Fluctuations in InGaN Light Emitting Diodes

The accurate determination of the compositional fluctuations is pivotal in understanding their role in the reduction of efficiency in high indium content $In_{x}Ga_{1-x}N$ light-emitting diodes, the origin of which is still poorly understood. Here we have combined electron energy loss spectroscopy (EELS) imaging at sub-nanometer resolution with multiscale computational models to obtain a statistical distribution of the compositional fluctuations in $In_{x}Ga_{1-x}N$ quantum wells (QWs). Employing a multiscale computational model, we show the tendency of intrinsic compositional fluctuation in $In_{x}Ga_{1-x}N$ QWs at different Indium concentration and in the presence of strain. We have developed a systematic formalism based on the autonomous detection of compositional fluctuation in observed and simulated EELS maps. We have shown a direct comparison between the computationally predicted and experimentally observed compositional fluctuations. We have found that although a random alloy model captures the distribution of compositional fluctuations in relatively low In ($\sim$ 18%) content $In_{x}Ga_{1-x}N$ QWs, there exists a striking deviation from the model in higher In content ($\geq$ 24%) QWs. Our results highlight a distinct behavior in carrier localization driven by compositional fluctuations in the low and high In-content InGaN QWs, which would ultimately affect the performance of LEDs. Furthermore, our robust computational and atomic characterization method can be widely applied to study materials in which nanoscale compositional fluctuations play a significant role on the material performance.

cond-mat.mtrl-sci↗

arXiv2021

A Survey on Concept Factorization: From Shallow to Deep Representation Learning

The quality of learned features by representation learning determines the performance of learning algorithms and the related application tasks (such as high-dimensional data clustering). As a relatively new paradigm for representation learning, Concept Factorization (CF) has attracted a great deal of interests in the areas of machine learning and data mining for over a decade. Lots of effective CF based methods have been proposed based on different perspectives and properties, but note that it still remains not easy to grasp the essential connections and figure out the underlying explanatory factors from exiting studies. In this paper, we therefore survey the recent advances on CF methodologies and the potential benchmarks by categorizing and summarizing the current methods. Specifically, we first re-view the root CF method, and then explore the advancement of CF-based representation learning ranging from shallow to deep/multilayer cases. We also introduce the potential application areas of CF-based methods. Finally, we point out some future directions for studying the CF-based representation learning. Overall, this survey provides an insightful overview of both theoretical basis and current developments in the field of CF, which can also help the interested researchers to understand the current trends of CF and find the most appropriate CF techniques to deal with particular applications.

cs.LG↗

arXiv2021

Failure Prediction in Production Line Based on Federated Learning: An Empirical Study

Data protection across organizations is limiting the application of centralized learning (CL) techniques. Federated learning (FL) enables multiple participants to build a learning model without sharing data. Nevertheless, there are very few research works on FL in intelligent manufacturing. This paper presents the results of an empirical study on failure prediction in the production line based on FL. This paper (1) designs Federated Support Vector Machine (FedSVM) and Federated Random Forest (FedRF) algorithms for the horizontal FL and vertical FL scenarios, respectively; (2) proposes an experiment process for evaluating the effectiveness between the FL and CL algorithms; (3) finds that the performance of FL and CL are not significantly different on the global testing data, on the random partial testing data, and on the estimated unknown Bosch data, respectively. The fact that the testing data is heterogeneous enhances our findings. Our study reveals that FL can replace CL for failure prediction.

cs.LG↗

arXiv2021

Surface-acoustic-wave-controlled optomechanically induced transparency in a hybrid piezo-optomechanical planar distributed Bragg-reflector-cavity system

We propose a scheme that can generate tunable optomechanically induced transparency in a hybrid piezooptomechanical cavity system. The system is composed of a high-quality planar distributed Bragg reflector cavity modified with an embedded Gaussian-shaped defect. Moreover, interdigitated transducers are fabricated on the surface of the cavity to generate surface acoustic waves. Under the actuation of the surface acoustic wave, the upper Bragg mirrors can be vibrated as a bulk acoustic resonator, and the distributed Bragg reflector cavity becomes a standard three-level optomechanical system. In this situation, we show that when a strong pump optical field and a weak probe optical field are simultaneously applied to the hybrid optomechanical cavity system, optomechanically induced transparency occurs under the quantum interference between different energy-level pathways. Our scheme can be applied in the fields of optical switches and quantum information processing in solid-state quantum systems.

quant-ph↗

arXiv2020

Sub-sampled Cross-component Prediction for Emerging Video Coding Standards

Cross-component linear model (CCLM) prediction has been repeatedly proven to be effective in reducing the inter-channel redundancies in video compression. Essentially speaking, the linear model is identically trained by employing accessible luma and chroma reference samples at both encoder and decoder, elevating the level of operational complexity due to the least square regression or max-min based model parameter derivation. In this paper, we investigate the capability of the linear model in the context of sub-sampled based cross-component correlation mining, as a means of significantly releasing the operation burden and facilitating the hardware and software design for both encoder and decoder. In particular, the sub-sampling ratios and positions are elaborately designed by exploiting the spatial correlation and the inter-channel correlation. Extensive experiments verify that the proposed method is characterized by its simplicity in operation and robustness in terms of rate-distortion performance, leading to the adoption by Versatile Video Coding (VVC) standard and the third generation of Audio Video Coding Standard (AVS3).

cs.MM↗

arXiv2020

The Analog Front-end for the LGAD Based Precision Timing Application in CMS ETL

The analog front-end for the Low Gain Avalanche Detector (LGAD) based precision timing application in the CMS Endcap Timing Layer (ETL) has been prototyped in a 65 nm CMOS mini-ASIC named ETROC0. Serving as the very first prototype of ETL readout chip (ETROC), ETROC0 aims to study and demonstrate the performance of the analog frontend, with the goal to achieve 40 to 50 ps time resolution per hit with LGAD (therefore reach about 30ps per track with two detector-layer hits per track). ETROC0 consists of preamplifier and discriminator stages, which amplifies the LGAD signal and generates digital pulses containing time of arrival and time over threshold information. This paper will focus on the design considerations that lead to the ETROC front-end architecture choice, the key design features of the building blocks, the methodology of using the LGAD simulation data to evaluate and optimize the front-end design. The ETROC0 prototype chips have been extensively tested using charge injection and the measured performance agrees well with simulation. The initial beam test results are also presented, with time resolution of around 33 ps observed from the preamplifier waveform analysis and around 41 ps from the discriminator pulses analysis. A subset of ETROC0 chips have also been tested to a total ionizing dose of 100 MRad with X-ray and no performance degradation been observed.

physics.ins-det↗

arXiv2020

Unifying Homophily and Heterophily Network Transformation via Motifs

Higher-order proximity (HOP) is fundamental for most network embedding methods due to its significant effects on the quality of node embedding and performance on downstream network analysis tasks. Most existing HOP definitions are based on either homophily to place close and highly interconnected nodes tightly in embedding space or heterophily to place distant but structurally similar nodes together after embedding. In real-world networks, both can co-exist, and thus considering only one could limit the prediction performance and interpretability. However, there is no general and universal solution that takes both into consideration. In this paper, we propose such a simple yet powerful framework called homophily and heterophliy preserving network transformation (H2NT) to capture HOP that flexibly unifies homophily and heterophily. Specifically, H2NT utilises motif representations to transform a network into a new network with a hybrid assumption via micro-level and macro-level walk paths. H2NT can be used as an enhancer to be integrated with any existing network embedding methods without requiring any changes to latter methods. Because H2NT can sparsify networks with motif structures, it can also improve the computational efficiency of existing network embedding methods when integrated. We conduct experiments on node classification, structural role classification and motif prediction to show the superior prediction performance and computational efficiency over state-of-the-art methods. In particular, DeepWalk-based H2 NT achieves 24% improvement in terms of precision on motif prediction, while reducing 46% computational time compared to the original DeepWalk.

cs.SI↗

arXiv2020

Hop-Hop Relation-aware Graph Neural Networks

Graph Neural Networks (GNNs) are widely used in graph representation learning. However, most GNN methods are designed for either homogeneous or heterogeneous graphs. In this paper, we propose a new model, Hop-Hop Relation-aware Graph Neural Network (HHR-GNN), to unify representation learning for these two types of graphs. HHR-GNN learns a personalized receptive field for each node by leveraging knowledge graph embedding to learn relation scores between the central node's representations at different hops. In neighborhood aggregation, our model simultaneously allows for hop-aware projection and aggregation. This mechanism enables the central node to learn a hop-wise neighborhood mixing that can be applied to both homogeneous and heterogeneous graphs. Experimental results on five benchmarks show the competitive performance of our model compared to state-of-the-art GNNs, e.g., up to 13K faster in terms of time cost per training epoch on large heterogeneous graphs.

cs.LG↗

arXiv2020

Intent Detection with WikiHow

Modern task-oriented dialog systems need to reliably understand users' intents. Intent detection is most challenging when moving to new domains or new languages, since there is little annotated data. To address this challenge, we present a suite of pretrained intent detection models. Our models are able to predict a broad range of intended goals from many actions because they are trained on wikiHow, a comprehensive instructional website. Our models achieve state-of-the-art results on the Snips dataset, the Schema-Guided Dialogue dataset, and all 3 languages of the Facebook multilingual dialog datasets. Our models also demonstrate strong zero- and few-shot performance, reaching over 75% accuracy using only 100 training examples in all datasets.

cs.CL↗

arXiv2020

Reasoning about Goals, Steps, and Temporal Ordering with WikiHow

We propose a suite of reasoning tasks on two types of relations between procedural events: goal-step relations ("learn poses" is a step in the larger goal of "doing yoga") and step-step temporal relations ("buy a yoga mat" typically precedes "learn poses"). We introduce a dataset targeting these two relations based on wikiHow, a website of instructional how-to articles. Our human-validated test set serves as a reliable benchmark for commonsense inference, with a gap of about 10% to 20% between the performance of state-of-the-art transformer models and human performance. Our automatically-generated training set allows models to effectively transfer to out-of-domain tasks requiring knowledge of procedural events, with greatly improved performances on SWAG, Snips, and the Story Cloze Test in zero- and few-shot settings.

cs.CL↗

arXiv2020

Small but Mighty: New Benchmarks for Split and Rephrase

Split and Rephrase is a text simplification task of rewriting a complex sentence into simpler ones. As a relatively new task, it is paramount to ensure the soundness of its evaluation benchmark and metric. We find that the widely used benchmark dataset universally contains easily exploitable syntactic cues caused by its automatic generation process. Taking advantage of such cues, we show that even a simple rule-based model can perform on par with the state-of-the-art model. To remedy such limitations, we collect and release two crowdsourced benchmark datasets. We not only make sure that they contain significantly more diverse syntax, but also carefully control for their quality according to a well-defined set of criteria. While no satisfactory automatic metric exists, we apply fine-grained manual evaluation based on these criteria using crowdsourcing, showing that our datasets better represent the task and are significantly more challenging for the models.

cs.CL↗