arXiv subjects

Xin Zheng

Publications and source records attributed to Xin Zheng.

89 records · Page 5Linked to original sources

arXiv2021

Robust Physical-World Attacks on Face Recognition

Face recognition has been greatly facilitated by the development of deep neural networks (DNNs) and has been widely applied to many safety-critical applications. However, recent studies have shown that DNNs are very vulnerable to adversarial examples, raising serious concerns on the security of real-world face recognition. In this work, we study sticker-based physical attacks on face recognition for better understanding its adversarial robustness. To this end, we first analyze in-depth the complicated physical-world conditions confronted by attacking face recognition, including the different variations of stickers, faces, and environmental conditions. Then, we propose a novel robust physical attack framework, dubbed PadvFace, to model these challenging variations specifically. Furthermore, considering the difference in attack complexity, we propose an efficient Curriculum Adversarial Attack (CAA) algorithm that gradually adapts adversarial stickers to environmental variations from easy to complex. Finally, we construct a standardized testing protocol to facilitate the fair evaluation of physical attacks on face recognition, and extensive experiments on both dodging and impersonation attacks demonstrate the superior performance of the proposed method.

cs.CV↗

arXiv2021

Adaptive Nearest Neighbor Machine Translation

kNN-MT, recently proposed by Khandelwal et al. (2020a), successfully combines pre-trained neural machine translation (NMT) model with token-level k-nearest-neighbor (kNN) retrieval to improve the translation accuracy. However, the traditional kNN algorithm used in kNN-MT simply retrieves a same number of nearest neighbors for each target token, which may cause prediction errors when the retrieved neighbors include noises. In this paper, we propose Adaptive kNN-MT to dynamically determine the number of k for each target token. We achieve this by introducing a light-weight Meta-k Network, which can be efficiently trained with only a few training samples. On four benchmark machine translation datasets, we demonstrate that the proposed method is able to effectively filter out the noises in retrieval results and significantly outperforms the vanilla kNN-MT model. Even more noteworthy is that the Meta-k Network learned on one domain could be directly applied to other domains and obtain consistent improvements, illustrating the generality of our method. Our implementation is open-sourced at https://github.com/zhengxxn/adaptive-knn-mt.

cs.CL↗

arXiv2021

Efficient LiDAR Odometry for Autonomous Driving

LiDAR odometry plays an important role in self-localization and mapping for autonomous navigation, which is usually treated as a scan registration problem. Although having achieved promising performance on KITTI odometry benchmark, the conventional searching tree-based approach still has the difficulty in dealing with the large scale point cloud efficiently. The recent spherical range image-based method enjoys the merits of fast nearest neighbor search by spherical mapping. However, it is not very effective to deal with the ground points nearly parallel to LiDAR beams. To address these issues, we propose a novel efficient LiDAR odometry approach by taking advantage of both non-ground spherical range image and bird's-eye-view map for ground points. Moreover, a range adaptive method is introduced to robustly estimate the local surface normal. Additionally, a very fast and memory-efficient model update scheme is proposed to fuse the points and their corresponding normals at different time-stamps. We have conducted extensive experiments on KITTI odometry benchmark, whose promising results demonstrate that our proposed approach is effective.

cs.RO↗

arXiv2021

HypoNLI: Exploring the Artificial Patterns of Hypothesis-only Bias in Natural Language Inference

Many recent studies have shown that for models trained on datasets for natural language inference (NLI), it is possible to make correct predictions by merely looking at the hypothesis while completely ignoring the premise. In this work, we manage to derive adversarial examples in terms of the hypothesis-only bias and explore eligible ways to mitigate such bias. Specifically, we extract various phrases from the hypotheses (artificial patterns) in the training sets, and show that they have been strong indicators to the specific labels. We then figure out `hard' and `easy' instances from the original test sets whose labels are opposite to or consistent with those indications. We also set up baselines including both pretrained models (BERT, RoBERTa, XLNet) and competitive non-pretrained models (InferSent, DAM, ESIM). Apart from the benchmark and baselines, we also investigate two debiasing approaches which exploit the artificial pattern modeling to mitigate such hypothesis-only bias: down-sampling and adversarial training. We believe those methods can be treated as competitive baselines in NLI debiasing tasks.

cs.CL↗

arXiv2021

Towards Faithfulness in Open Domain Table-to-text Generation from an Entity-centric View

In open domain table-to-text generation, we notice that the unfaithful generation usually contains hallucinated content which can not be aligned to any input table record. We thus try to evaluate the generation faithfulness with two entity-centric metrics: table record coverage and the ratio of hallucinated entities in text, both of which are shown to have strong agreement with human judgements. Then based on these metrics, we quantitatively analyze the correlation between training data quality and generation fidelity which indicates the potential usage of entity information in faithful generation. Motivated by these findings, we propose two methods for faithful generation: 1) augmented training by incorporating the auxiliary entity information, including both an augmented plan-based model and an unsupervised model and 2) training instance selection based on faithfulness ranking. We show these approaches improve generation fidelity in both full dataset setting and few shot learning settings by both automatic and human evaluations.

cs.CL↗

arXiv2020

An Empirical Study on Model-agnostic Debiasing Strategies for Robust Natural Language Inference

The prior work on natural language inference (NLI) debiasing mainly targets at one or few known biases while not necessarily making the models more robust. In this paper, we focus on the model-agnostic debiasing strategies and explore how to (or is it possible to) make the NLI models robust to multiple distinct adversarial attacks while keeping or even strengthening the models' generalization power. We firstly benchmark prevailing neural NLI models including pretrained ones on various adversarial datasets. We then try to combat distinct known biases by modifying a mixture of experts (MoE) ensemble method and show that it's nontrivial to mitigate multiple NLI biases at the same time, and that model-level ensemble method outperforms MoE ensemble method. We also perform data augmentation including text swap, word substitution and paraphrase and prove its efficiency in combating various (though not all) adversarial attacks at the same time. Finally, we investigate several methods to merge heterogeneous training data (1.35M) and perform model ensembling, which are straightforward but effective to strengthen NLI models.

cs.CL↗

arXiv2020

Impacts of random filling on spin squeezing via Rydberg dressing in optical clocks

We analyze spin squeezing via Rydberg dressing in optical lattice clocks with random fractional filling. We compare the achievable clock stability in different lattice geometries, including unity-filled tweezer clock arrays and fractionally filled lattice clocks with varying dimensionality. We provide practical considerations and useful tools in the form of approximate analytical expressions and fitting functions to aid in the experimental implementation of Rydberg-dressed spin squeezing. We demonstrate that spin squeezing via Rydberg dressing in one-, two-, and three-dimensional optical lattices can provide significant improvements in stability in the presence of random fractional filling.

quant-ph↗

arXiv2020

Quantile Treatment Effects and Bootstrap Inference under Covariate-Adaptive Randomization

In this paper, we study the estimation and inference of the quantile treatment effect under covariate-adaptive randomization. We propose two estimation methods: (1) the simple quantile regression and (2) the inverse propensity score weighted quantile regression. For the two estimators, we derive their asymptotic distributions uniformly over a compact set of quantile indexes, and show that, when the treatment assignment rule does not achieve strong balance, the inverse propensity score weighted estimator has a smaller asymptotic variance than the simple quantile regression estimator. For the inference of method (1), we show that the Wald test using a weighted bootstrap standard error under-rejects. But for method (2), its asymptotic size equals the nominal level. We also show that, for both methods, the asymptotic size of the Wald test using a covariate-adaptive bootstrap standard error equals the nominal level. We illustrate the finite sample performance of the new estimation and inference methods using both simulated and real datasets.

stat.ME↗

arXiv2020

Two-dimensional Multi-fiber Spectrum Image Correction Based on Machine Learning Techniques

Due to limited size and imperfect of the optical components in a spectrometer, aberration has inevitably been brought into two-dimensional multi-fiber spectrum image in LAMOST, which leads to obvious spacial variation of the point spread functions (PSFs). Consequently, if spatial variant PSFs are estimated directly , the huge storage and intensive computation requirements result in deconvolutional spectral extraction method become intractable. In this paper, we proposed a novel method to solve the problem of spatial variation PSF through image aberration correction. When CCD image aberration is corrected, PSF, the convolution kernel, can be approximated by one spatial invariant PSF only. Specifically, machine learning techniques are adopted to calibrate distorted spectral image, including Total Least Squares (TLS) algorithm, intelligent sampling method, multi-layer feed-forward neural networks. The calibration experiments on the LAMOST CCD images show that the calibration effect of proposed method is effectible. At the same time, the spectrum extraction results before and after calibration are compared, results show the characteristics of the extracted one-dimensional waveform are more close to an ideal optics system, and the PSF of the corrected object spectrum image estimated by the blind deconvolution method is nearly central symmetry, which indicates that our proposed method can significantly reduce the complexity of spectrum extraction and improve extraction accuracy.

astro-ph.IM↗

arXiv2019

A Survey of Deep Facial Attribute Analysis

Facial attribute analysis has received considerable attention when deep learning techniques made remarkable breakthroughs in this field over the past few years. Deep learning based facial attribute analysis consists of two basic sub-issues: facial attribute estimation (FAE), which recognizes whether facial attributes are present in given images, and facial attribute manipulation (FAM), which synthesizes or removes desired facial attributes. In this paper, we provide a comprehensive survey of deep facial attribute analysis from the perspectives of both estimation and manipulation. First, we summarize a general pipeline that deep facial attribute analysis follows, which comprises two stages: data preprocessing and model construction. Additionally, we introduce the underlying theories of this two-stage pipeline for both FAE and FAM. Second, the datasets and performance metrics commonly used in facial attribute analysis are presented. Third, we create a taxonomy of state-of-the-art methods and review deep FAE and FAM algorithms in detail. Furthermore, several additional facial attribute related issues are introduced, as well as relevant real-world applications. Finally, we discuss possible challenges and promising future research directions.

cs.CV↗

arXiv2019

Pulsar Candidate Selection Using Ensemble Networks for FAST Drift-Scan Survey

The Commensal Radio Astronomy Five-hundred-meter Aperture Spherical radio Telescope (FAST) Survey (CRAFTS) utilizes the novel drift-scan commensal survey mode of FAST and can generate billions of pulsar candidate signals. The human experts are not likely to thoroughly examine these signals, and various machine sorting methods are used to aid the classification of the FAST candidates. In this study, we propose a new ensemble classification system for pulsar candidates. This system denotes the further development of the pulsar image-based classification system (PICS), which was used in the Arecibo Telescope pulsar survey, and has been retrained and customized for the FAST drift-scan survey. In this study, we designed a residual network model comprising 15 layers to replace the convolutional neural networks (CNNs) in PICS. The results of this study demonstrate that the new model can sort >96% of real pulsars to belong the top 1% of all candidates and classify >1.6 million candidates per day using a dual--GPU and 24--core computer. This increased speed and efficiency can help to facilitate real-time or quasi-real-time processing of the pulsar-search data stream obtained from CRAFTS. In addition, we have published the labeled FAST data used in this study online, which can aid in the development of new deep learning techniques for performing pulsar searches.

astro-ph.IM↗

arXiv2018

A Survey of Location Prediction on Twitter

Locations, e.g., countries, states, cities, and point-of-interests, are central to news, emergency events, and people's daily lives. Automatic identification of locations associated with or mentioned in documents has been explored for decades. As one of the most popular online social network platforms, Twitter has attracted a large number of users who send millions of tweets on daily basis. Due to the world-wide coverage of its users and real-time freshness of tweets, location prediction on Twitter has gained significant attention in recent years. Research efforts are spent on dealing with new challenges and opportunities brought by the noisy, short, and context-rich nature of tweets. In this survey, we aim at offering an overall picture of location prediction on Twitter. Specifically, we concentrate on the prediction of user home locations, tweet locations, and mentioned locations. We first define the three tasks and review the evaluation metrics. By summarizing Twitter network, tweet content, and tweet context as potential inputs, we then structurally highlight how the problems depend on these inputs. Each dependency is illustrated by a comprehensive review of the corresponding strategies adopted in state-of-the-art approaches. In addition, we also briefly review two related problems, i.e., semantic location prediction and point-of-interest recommendation. Finally, we list future research directions.

cs.SI↗

arXiv2017

Character-Based Text Classification using Top Down Semantic Model for Sentence Representation

Despite the success of deep learning on many fronts especially image and speech, its application in text classification often is still not as good as a simple linear SVM on n-gram TF-IDF representation especially for smaller datasets. Deep learning tends to emphasize on sentence level semantics when learning a representation with models like recurrent neural network or recursive neural network, however from the success of TF-IDF representation, it seems a bag-of-words type of representation has its strength. Taking advantage of both representions, we present a model known as TDSM (Top Down Semantic Model) for extracting a sentence representation that considers both the word-level semantics by linearly combining the words with attention weights and the sentence-level semantics with BiLSTM and use it on text classification. We apply the model on characters and our results show that our model is better than all the other character-based and word-based convolutional neural network models by \cite{zhang15} across seven different datasets with only 1\% of their parameters. We also demonstrate that this model beats traditional linear models on TF-IDF vectors on small and polished datasets like news article in which typically deep learning models surrender.

cs.CL↗

arXiv2015

Application of Deep Neural Network in Estimation of the Weld Bead Parameters

We present a deep learning approach to estimation of the bead parameters in welding tasks. Our model is based on a four-hidden-layer neural network architecture. More specifically, the first three hidden layers of this architecture utilize Sigmoid function to produce their respective intermediate outputs. On the other hand, the last hidden layer uses a linear transformation to generate the final output of this architecture. This transforms our deep network architecture from a classifier to a non-linear regression model. We compare the performance of our deep network with a selected number of results in the literature to show a considerable improvement in reducing the errors in estimation of these values. Furthermore, we show its scalability on estimating the weld bead parameters with same level of accuracy on combination of datasets that pertain to different welding techniques. This is a nontrivial result that is counter-intuitive to the general belief in this field of research.

cs.LG↗

arXiv2015

Mid-infrared ultra-short mode-locked fiber laser utilizing topological insulator Bi2Te3 nano-sheets as the saturable absorber

The newly-emergent two-dimensional topological insulators (TIs) have shown their unique electronic and optical properties, such as good thermal management, high nonlinear refraction index and ultrafast relaxation time. Their narrow energy band gaps predict their optical absorption ability further into the mid-infrared region and their possibility to be very broadband light modulators ranging from the visible to the mid-infrared region. In this paper, a mid-infrared mode-locked fluoride fiber laser with TI Bi2Te3 nano-sheets as the saturable absorber is presented. Continuous wave lasing, Q-switched and continuous-wave mode-locking (CW-ML) operations of the laser are observed sequentially by increasing the pump power. The observed CW-ML pulse train has a pulse repetition rate of 10.4 MHz, a pulse width of ~6 ps, and a center wavelength of 2830 nm. The maximum achievable pulse energy is 8.6 nJ with average power up to 90 mW. This work forcefully demonstrates the promising applications of two-dimensional TIs for ultra-short laser operation and nonlinear optics in the mid-infrared region.

physics.optics↗

arXiv2015

Black phosphorus as a new broadband saturable absorber for infrared passively Q-switched fiber lasers

Black phosphorus (BP) with its enticing electric and optical properties is intensely researched in the field of optoelectronics. In this paper, Q-switched pulses at 1550 nm and 2 um wavelengths are obtained by inserting bulk-structured BP based saturable absorber (SA) into an erbium-doped fiber laser (EDFL) and an thulium/holmium-doped fiber laser (THDFL), respectively. The BP-SA was prepared by depositing powered BP material on to the flat side of a side-polished single mode fiber. Q-switched 1550 nm pulses with width tuned from 9.35 to 31 us were obtained for the EDFL. For the THDFL, over 100 nm wavelength range could be achieved from 1832 to 1935 nm by adjusting the pump power. To the best of our knowledge, these results demonstrated the broadband saturable absorption property of BP and for the first time verified that BP as a new two-dimensional material for applications in saturable absorption devices.

physics.optics↗

arXiv2013

Feature Learning with Gaussian Restricted Boltzmann Machine for Robust Speech Recognition

In this paper, we first present a new variant of Gaussian restricted Boltzmann machine (GRBM) called multivariate Gaussian restricted Boltzmann machine (MGRBM), with its definition and learning algorithm. Then we propose using a learned GRBM or MGRBM to extract better features for robust speech recognition. Our experiments on Aurora2 show that both GRBM-extracted and MGRBM-extracted feature performs much better than Mel-frequency cepstral coefficient (MFCC) with either HMM-GMM or hybrid HMM-deep neural network (DNN) acoustic model, and MGRBM-extracted feature is slightly better.

cs.CL↗