检索结果-内蒙古大学图书馆

Efficient sentence embedding via semantic subspace analysis

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Wang, Bin Chen, Fenxiao Wang, Yuncheng Jay Kuo, C.-C. Signal and Image Processing Institute Department of Electrical and Computer Engineering University of Southern California Los AngelesCA United States

A novel sentence embedding method built upon semantic subspace analysis, called semantic subspace sentence embedding (S3E), is proposed in this work. Given the fact that word embeddings can capture semantic relationship while semantically similar words tend to form semantic groups in a high-dimensional embedding space, we develop a sentence representation scheme by analyzing semantic subspaces of its constituent words. Specifically, we construct a sentence model from two aspects. First, we represent words that lie in the same semantic group using the intra-group descriptor. Second, we characterize the interaction between multiple semantic groups with the inter-group descriptor. The proposed S3E method is evaluated on both textual similarity tasks and supervised tasks. Experimental results show that it offers comparable or better performance than the state-of-the-art. The complexity of our S3E method is also much lower than other parameterized models. Copyright © 2020, The Authors. All rights reserved.

关键词： Embeddings

Asymmetry Total Variation and Framelet Regularized Nonconvex Low-Rank Tensor Completion

学校读者我要写书评

暂无评论

SSRN

SSRN 2022年

作者： Chen, Yongyong Zhao, Xiaojia Xu, Tingting Zeng, Haijin Xu, Yanhui Chen, Junxing The School of Computer Science and Technology Harbin Institute of Technology Shenzhen518055 China Guangdong Provincial Key Laboratory of Novel Security Intelligence Technologies Shenzhen518055 China Image Processing and Interpretation Imec Research Group Ghent University Belgium Department of Electrical and Computer Engineering Faculty of Science and Technology University of Macau China The College of Medicine and Biological Information Engineering Northeastern University Shenyang110004 China

The low-rank tensor representation has shown enormous potential and advantages in diverse applications. However, (1) due to the inherent limitations of the low-rank tensor model, the tensor nuclear norm is utilized yet as a biased estimate of the tensor rank, which limits the image recovery performance. (2) most current low-rank approximation-based methods focus on the preservation of global information, while ignoring the local details preservation such as the spatial piece-wise smoothness and sharp edges. To overcome he above obstacles, we propose a novel asymmetry three-dimensional total variation and framelet regularized nonconvex low-rank tensor completion (ATV-FNTC) model, which integrates the nonconvex penalty function and asymmetric three-dimensional total variation into one unified model. Specifically, ATV-FNTC introduces one nonconvex penalty function as a tighter regularizer to approximate the tensor rank. Different from existing methods, the asymmetric three-dimensional total variation regularization is developed to achieve more accurate recovery of detailed information and flexibly control the smoothing strength of different dimensions of tensor data. Furthermore, we design an iterative algorithm to solve the nonconvex ATV-FNTC model by the alternating direction methods of multipliers. Experimental results reveal the superiority of the proposed method compared with several state-of-the-art methods when handling different completion tasks. © 2022, The Authors. All rights reserved.

关键词： Tensors

SBERT-WK: A sentence embedding method by dissecting BERT-based word models

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Wang, Bin Jay Kuo, C.-C. Ming-Hsieh Department of Electrical and Computer Engineering and Signal Image Processing Institute University of Southern California Los AngelesCA90089-2584 United States Ming-Hsieh Department of Electrical and Computer Engineering Department of Computer Science and Signal Image Processing Institute University of Southern California Los AngelesCA90089-2584 United States

Sentence embedding is an important research topic in natural language processing (NLP) since it can transfer knowledge to downstream tasks. Meanwhile, a contextualized word representation, called BERT, achieves the state-of-the-art performance in quite a few NLP tasks. Yet, it is an open problem to generate a high quality sentence representation from BERTbased word models. It was shown in previous study that different layers of BERT capture different linguistic properties. This allows us to fusion information across layers to find better sentence representation. In this work, we study the layer-wise pattern of the word representation of deep contextualized models. Then, we propose a new sentence embedding method by dissecting BERT-based word models through geometric analysis of the space spanned by the word representation. It is called the SBERT-WK method. No further training is required in SBERTWK. We evaluate SBERT-WK on semantic textual similarity and downstream supervised tasks. Furthermore, ten sentencelevel probing tasks are presented for detailed linguistic analysis. Experiments show that SBERT-WK achieves the state-of-the-art performance. Our codes are publicly available at the Github1. Copyright © 2020, The Authors. All rights reserved.

关键词： Embeddings

Twenty-Five Years of Advances in Beamforming: From Convex and Nonconvex Optimization to Learning Techniques

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Elbir, Ahmet M. Mishra, Kumar Vijay Vorobyov, Sergiy A. Heath, Robert W. The Interdisciplinary Centre for Security Reliability and Trust University of Luxembourg Luxembourg Duzce University Duzce Turkey The United States DEVCOM Army Research Laboratory AdelphiMD20783 United States The Department of Signal Processing and Acoustics Aalto University Espoo02150 Finland The Department of Electrical and Computer Engineering North Carolina State University RaleighNC27606 United States

Beamforming is a signal processing technique to steer, shape, and focus an electromagnetic wave using an array of sensors toward a desired direction. It has been used in several engineering applications such as radar, sonar, acoustics, astronomy, seismology, medical imaging, and communications. With the advent of multi-antenna technologies in, say, radar and communication, there has been a great interest in designing beamformers by exploiting convex or nonconvex optimization methods. Recently, machine learning is also leveraged for obtaining attractive solutions to more complex beamforming scenarios. This article captures the evolution of beamforming in the last twenty-five years from convex-to-nonconvex optimization and optimization-to-learning approaches. It provides a glimpse into these important signal processing algorithms for a variety of transmit-receive architectures, propagation zones, propagation paths, and multi-disciplinary applications. Copyright © 2022, The Authors. All rights reserved.

关键词： Beamforming

Physics-Based Optical Flow Estimation Under Varying Illumination Conditions

学校读者我要写书评

暂无评论

SSRN

SSRN 2022年

作者： Liao, Xiaoxin Cai, Zemin Chen, Jun Liu, Tianshu Lai, Jian-Huang The Department of Electronic Engineering Shantou University Guangdong Shantou515063 China Key Laboratory of Digital Signal and Image Processing of Guangdong Province Guangdong Shantou515063 China The School of Industrial Design and Ceramic Art Foshan University Foshan528000 China The Department of Mechanical and Aerospace Engineering Western Michigan University KalamazooMI49008 United States The School of Data and Computer Science Sun Yat-sen University Guangdong Guangzhou510006 China

The physics-based optical flow (PBOF) model was derived from typical flow visulizations and it provided a more solid physical foundation for optical flow calculation. However, like many recent variational optical flow methods, it is not formulated to deal with the effect of illumination variation. In this paper, we propose a new efficient illumination-invariant optical flow estimation method, which improves the physics-based optical flow model. The proposed method introduces a multiplicative product of the velocity gradient and intensity from PBOF and an additive diffusion component to relax the brightness constancy assumption. The proposed method employs a linear transformation to reduce the influence of varying illumination on the data term. Then, a smoothness-sparsity regularization is used to constrain the minimization problem to have good edge-preserving for optical flow. Furthermore, the numerical implementation is given to solve the optimization problem. The accuracy of the developed method and the robustness against varying illumination are evaluated. The proposed method can provide motion estimation with acceptable accuracy for three optical flow datasets and cloud images of Jupiter’s great red spot, which have challenging illumination variations. © 2022, The Authors. All rights reserved.

关键词： Optical flows

Weakly-Supervised Semantic Segmentation of Circular-Scan, Synthetic-Aperture-Sonar imagery

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Sledge, Isaac J. Byrne, Dominic M. King, Jonathan L. Ostertag, Steven H. Woods, Denton L. Prater, James L. Kennedy, Jermaine L. Marston, Timothy M. Príncipe, José C. The Advanced Signal Processing and Automated Target Recognition Branch Naval Surface Warfare Center Panama CityFL United States The Naval Sea Systems Command United States The Applied Sensing and Processing Branch Naval Surface Warfare Center Panama CityFL United States The Littoral Acoustics and Target Physics Branch Naval Surface Warfare Center Panama CityFL United States The Division Head of the Sensing Sciences and Systems Division Naval Surface Warfare Center Panama CityFL United States The Applied Physics Laboratory University of Washington SeattleWA United States The Department of Electrical and Computer Engineering The Department of Biomedical Engineering University of Florida GainesvilleFL United States The University of Florida United States

We propose a weakly-supervised framework for the semantic segmentation of circular-scan synthetic-aperture-sonar (CSAS) imagery. The first part of our framework is trained in a supervised manner, on image-level labels, to uncover a set of semi-sparse, spatially-discriminative regions in each image. The classification uncertainty of each region is then evaluated. Those areas with the lowest uncertainties are then chosen to be weakly labeled segmentation seeds, at the pixel level, for the second part of the framework. Each of the seed extents are progressively resized according to an unsupervised, information-theoretic loss with structured-prediction regularizers. This reshaping process uses multi-scale, adaptively-weighted features to delineate class-specific transitions in local image content. Content-addressable memories are inserted at various parts of our framework so that it can leverage features from previously seen images to improve segmentation performance for related images. We evaluate our weakly-supervised framework using real-world CSAS imagery that contains over ten seafloor classes and ten target classes. We show that our framework performs comparably to nine fully-supervised deep networks. Our framework also outperforms eleven of the best weakly-supervised deep networks. We achieve state-of-the-art performance when pre-training on natural imagery. The average absolute performance gap to the next-best weakly-supervised network is well over ten percent for both natural imagery and sonar imagery. This gap is found to be statistically significant. © 2024, CC BY.

关键词： Semantic Segmentation

Attention-Guided Generative Adversarial Network for Whisper to Normal Speech Conversion

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Gao, Teng Zhou, Jian Wang, Huabin Tao, Liang Kwan, Hon Keung Key Laboratory of Intelligent Computing and Signal Processing Ministry of Education Anhui University Hefei230601 China Department of Electrical and Computer Engineering University of Windsor WindsorONN9B 3P4 Canada

Whispered speech is a special way of pronunciation without using vocal cord vibration. A whispered speech does not contain a fundamental frequency, and its energy is about 20dB lower than that of a normal speech. Converting a whispered speech into a normal speech can improve speech quality and intelligibility. In this paper, a novel attention-guided generative adversarial network model incorporating an autoencoder, a Siamese neural network, and an identity mapping loss function for whisper to normal speech conversion (AGANW2SC) is proposed. The proposed method avoids the challenge of estimating the fundamental frequency of the normal voiced speech converted from a whispered speech. Specifically, the proposed model is more amendable to practical applications because it does not need to align speech features for training. Experimental results demonstrate that the proposed AGANW2SC can obtain improved speech quality and intelligibility compared with dynamic-time-warping-based methods. Copyright © 2021, The Authors. All rights reserved.

关键词： Generative adversarial networks

CycleGAN with dual adversarial loss for bone-conducted speech enhancement

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Pan, Qing Gao, Teng Zhou, Jian Wang, Huabin Tao, Liang Kwan, Hon Keung Key Laboratory of Intelligent Computing and Signal Processing of Ministry of Education Anhui University Hefei230601 China Department of Electrical and Computer Engineering University of Windsor WindsorONN9B 3P4 Canada

Compared with air-conducted speech, bone-conducted speech has the unique advantage of shielding background noise. Enhancement of bone-conducted speech helps to improve its quality and intelligibility. In this paper, a novel CycleGAN with dual adversarial loss (CycleGAN-DAL) is proposed for bone-conducted speech enhancement. The proposed method uses an adversarial loss and a cycle-consistent loss simultaneously to learn forward and cyclic mapping, in which the adversarial loss is replaced with the classification adversarial loss and the defect adversarial loss to consolidate the forward mapping. Compared with conventional baseline methods, it can learn feature mapping between bone-conducted speech and target speech without additional air-conducted speech assistance. Moreover, the proposed method also avoids the over-smooth problem which is occurred commonly in conventional statistical based models. Experimental results show that the proposed method outperforms baseline methods such as CycleGAN, GMM, and BLSTM. © 2021, CC BY.

关键词： Speech enhancement

PALMNUT: An enhanced proximal alternating linearized minimization algorithm with application to separate regularization of magnitude and phase

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Liu, Yunsong Haldar, Justin P. Signal and Image Processing Institute Ming Hsieh Department of Electrical and Computer Engineering University of Southern California Los AngelesCA90089 United States

We introduce a new algorithm for complex image reconstruction with separate regularization of the image magnitude and phase. This optimization problem is interesting in many different image reconstruction contexts, although is nonconvex and can be difficult to solve. In this work, we first describe a novel implementation of the previous proximal alternating linearized minization (PALM) algorithm to solve this optimization problem. We then make enhancements to PALM, leading to a new algorithm named PALMNUT that combines the PALM together with Nesterov’s momentum and a novel approach that relies on uncoupled coordinatewise step sizes derived from coordinatewise Lipschitz-like bounds. Theoretically, we establish that a version of PALMNUT (without Nesterov’s momentum) monotonically decreases the objective function, leading to guaranteed convergence in many cases of interest. Empirical results obtained in the context of magnetic resonance imaging demonstrate that PALMNUT has computational advantages over common existing approaches like alternating minimization. Although our focus is on the application to separate magnitude and phase regularization, we expect that the same approach may also be useful in other nonconvex optimization problems with similar objective function structure. © 2020, CC BY-NC-SA.

关键词： Linearization