检索结果-内蒙古大学图书馆

Maximizing mutual information inside intra- and inter-modality for audio-visual event retrieval

INTERNATIONAL JOURNAL OF MULTIMEDIA INFORMATION RETRIEVAL 2023年第1期12卷 10页

作者： Li, Ruochen Li, Nannan Wang, Wenmin Macau Univ Sci & Technol Sch Engn & Comp Sci Ave Wai Long Taipa 999078 Macau Peoples R China

The human brain can process sound and visual information in overlapping areas of the cerebral cortex, which means that audio and visual information are deeply correlated with each other when we explore the world. To simulate this function of the human brain, audio-visual event retrieval (AVER) has been proposed. AVER is about using data from one modality (e.g., audio data) to query data from another. In this work, we aim to improve the performance of audio-visual event retrieval. To achieve this goal, first, we propose a novel network, InfoIIM, which enhance the accuracy of intra-model feature representation and inter-model feature alignment. The backbone of this network is a parallel connection of two VAE models with two different encoders and a shared decoder. Secondly, to enable the VAE to learn better feature representations and to improve intra-modal retrieval performance, we have used InfoMax-VAE instead of the vanilla VAE model. Additionally, we study the influence of modality-shared features on the effectiveness of audio-visual event retrieval. To verify the effectiveness of our proposed method, we validate our model on the AVE dataset, and the results show that our model outperforms several existing algorithms in most of the metrics. Finally, we present our future research directions, hoping to inspire relevant researchers.

关键词： Audio-visual retrieval variational autoencoder Mutual information InfoMax-VAE

来源：评论

学校读者我要写书评

暂无评论

A two-step framework for arbitrage-free prediction of the implied volatility surface

引用

QUANTITATIVE FINANCE 2023年第1期23卷 21-34页

作者： Zhang, Wenyong Li, Lingfei Zhang, Gongqiu Chinese Univ Hong Kong Dept Syst Engn & Engn Management Hong Kong Peoples R China Chinese Univ Hong Kong Sch Sci & Engn Shenzhen Peoples R China

In this study, we propose a two-step framework to predict the implied volatility surface (IVS) in a manner that excludes static arbitrage. First, we select features to represent the surface and predict them. Second, we use the predicted features to construct the IVS using a deep neural network (DNN) model by incorporating constraints that can prevent static arbitrage. We consider three methods to extract features from the implied volatility data: principal component analysis, variational autoencoder, and sampling the surface. We predict these features using the long short-term memory model. Additionally, we use a long time series of implied volatility data for S&P500 index options to train our models. We find that two feature construction methods (i.e. sampling the surface and variational autoencoders combined with DNN for surface construction) are the best performers in the out-of-sample prediction. Furthermore, both of them substantially outperform a popular regression model. We also find that the DNN model for surface construction not only removes static arbitrage but also significantly reduces the prediction error compared with a standard interpolation method.

关键词： Implied volatility surface Static arbitrage Prediction Deep learning variational autoencoder

来源：评论

学校读者我要写书评

暂无评论

Generating subsurface earth models using discrete representation learning and deep autoregressive network

引用

COMPUTATIONAL GEOSCIENCES 2023年第6期27卷 955-974页

作者： Chen, Jungang Huang, Chung-Kan Delgado, Jose F. Misra, Siddharth Texas A&M Univ Coll Engn Dept Petr Engn College Stn TX 77843 USA Conoco Phillips Houston TX USA Texas A&M Univ Coll Geosci Dept Geol & Geophys College Stn TX 77843 USA

Subsurface earth models (referred as geomodels) are crucial for characterizing complex subsurface systems. Multiple-point statistics is commonly used to generate geomodels. In this paper, a deep-learning-based generative method is developed as an alternative to traditional geomodel generation procedure. The generative method comprises two deep-learning models, namely hierarchical vector-quantized variational autoencoder (VQ-VAE-2) and PixelSNAIL autoregressive model. Based on the principle of neural discrete representation learning, the VQ-VAE-2 learns to massively compress the geomodels to extract the low-dimensional, discrete latent representation corresponding to each geomodel. Following that, PixelSNAIL uses deep autoregressive network to learn the prior distribution of the latent codes. For the purpose of geomodel generation, PixelSNAIL samples from the newly learnt prior distribution of latent codes, and then the decoder of the VQ-VAE-2 converts the newly sampled latent code to a newly constructed geomodel. PixelSNAIL can be used for unconditional or conditional geomodel generation. In unconditional generation, the generative workflow generates an ensemble of geomodels without any constraint. On the other hand, in the conditional geomodel generation, the generative workflow generates an ensemble of geomodels similar to a user-defined source image, which ultimately facilitates the control and manipulation of the generated geomodels. To better construct the fluvial channels in the geomodels, perceptual loss is implemented in the VQ-VAE-2 model instead of the traditional 'mean squared error' loss. At a specific compression ratio, the quality of multi-attribute geomodel generation is better than that of single-attribute geomodel generation.

关键词： Vector quantization variational autoencoder Perceptual loss Autoregression Deep generative model Compression

来源：评论

学校读者我要写书评

暂无评论

Non-linear Vibration Response Analysis of Rolling Bearing for Data Augmentation and Characterization

引用

JOURNAL OF VIBRATION ENGINEERING & TECHNOLOGIES 2023年第5期11卷 2109-2131页

作者： Rathore, Maan Singh Harsha, S. P. Indian Inst Technol Roorkee Dept Mech & Ind Engn Adv Mech Vibrat Lab Roorkee 247667 Uttarakhand India

The bearing fault diagnosis is carried out using nonlinear vibration responses using a proposed framework of VAEGAN-RDCNN. The condition of imbalanced data augmentation is solved using variational auto-encoder generative adversarial network (VAEGAN). The characterization of 2D patterns (phase space trajectories) into conditions of chaos (strong and weak attractor), and quasiperiodic is implemented using residual deep convolutional neural network (RDCNN). Experimentation is carried out on a test rig for phase space reconstruction. Various metrics such as PSNR, SSIM, KLD, and histogram analysis are used to investigate the quality of generated samples. Performance assessment of the proposed method is demonstrated using VAE, and WGAN models based on metrics namely average accuracy, precision, recall, and F1-score. The comparison results obtained are 98.20%, 90.13%, and 94.56% corresponding to VAEGAN-RDCNN, VAE, and WGAN, respectively. The discriminative feature learning ability of RDCNN is also presented using t-SNE technique. Hence, the proposed data augmentation and characterization methodology works effectively and achieves superior results to conventional data synthesis-based approaches.

关键词： Pattern classification Residual learning Generative adversarial networks variational autoencoder Deep convolutional neural network

来源：评论

学校读者我要写书评

暂无评论

scNAT: a deep learning method for integrating paired single-cell RNA and T cell receptor sequencing profiles

引用

GENOME BIOLOGY 2023年第1期24卷 1页

作者： Zhu, Biqing Wang, Yuge Ku, Li-Ting van Dijk, David Zhang, Le Hafler, David A. Zhao, Hongyu Yale Univ Program Computat Biol & Bioinformat New Haven CT 06511 USA Yale Univ Sch Publ Hlth Dept Biostat New Haven CT 06511 USA Yale Sch Med Dept Internal Med New Haven CT 06511 USA Yale Univ Dept Comp Sci New Haven CT 06511 USA Yale Univ Sch Med Dept Neurosci New Haven CT 06511 USA Yale Univ Sch Med Dept Neurol New Haven CT 06511 USA Yale Univ Sch Med Dept Immunobiol New Haven CT 06511 USA Aligning Sci Parkinsons ASAP Collaborat Res Networ Chevy Chase MD 20815 USA

Many deep learning-based methods have been proposed to handle complex single-cell data. Deep learning approaches may also prove useful to jointly analyze single-cell RNA sequencing (scRNA-seq) and single-cell T cell receptor sequencing (scTCR-seq) data for novel discoveries. We developed scNAT, a deep learning method that integrates paired scRNA-seq and scTCR-seq data to represent data in a unified latent space for downstream analysis. We demonstrate that scNAT is capable of removing batch effects, and identifying cell clusters and a T cell migration trajectory from blood to cerebrospinal fluid in multiple sclerosis.

关键词： scRNA-seq scTCR-seq Data integration Deep learning variational autoencoder Clone expansion

来源：评论

学校读者我要写书评

暂无评论

An Improved Convolutional Neural Network for Churn Analysis

引用

INTERNATIONAL JOURNAL OF ADVANCED COMPUTER SCIENCE AND APPLICATIONS 2023年第9期14卷 204-210页

作者： Gopal, Priya Bin MohdNawi, Nazri Univ Tun Hussein Onn Malaysia Fac Comp Sci & Informat Technol Parit Raja Malaysia

The significance of customer churn analysis has escalated due to the increasing availability of relevant data and intensifying competition. Researchers and practitioners are focused on enhancing prediction accuracy in modeling approaches, with deep neural networks emerging as appealing due to their robust performance across domains. However, the computational demands surge due to the challenges posed by dimensionality and inherent characteristics of the data. To address these issues, this research proposes a novel hybrid model that strategically integrates Convolutional Neural Networks (CNN) and a modified variational autoencoder (VAE). By carefully adjusting the parameters of the VAE to capture the central tendency and range of variation, the study aims to enhance the effectiveness of classifying high-dimensional churn data. The proposed framework's efficacy is evaluated using six benchmark datasets from various domains, with performance metrics encompassing accuracy, f1-score, precision, recall, and response time. Experimental results underscore the prowess of the hybrid technique in effectively handling high-dimensional and imbalanced time series data, thus offering a robust pathway for enhanced churn analysis.

关键词： Customer churn analysis deep learning variational autoencoder convolutional neural networks dimensionality reduction

来源：评论

学校读者我要写书评

暂无评论

New deep learning-based methods for visualizing ecosystem properties using environmental DNA metabarcoding data

引用

MOLECULAR ECOLOGY RESOURCES 2023年第8期23卷 1946-1958页

作者： Lamperti, Letizia Sanchez, Theophile Moussi, Sara Si Mouillot, David Albouy, Camille Fluck, Benjamin Bruno, Morgane Valentini, Alice Pellissier, Loic Manel, Stephanie PSL Univ Univ Montpellier EPHE CEFECNRSIRD Montpellier France Swiss Fed Inst Technol Dept Environm Syst Sci Ecosyst & Landscape Evolut Zurich Switzerland Swiss Fed Inst Forest Snow & Landscape Res WSL Land Change Sci Res Unit Ecosyst & Landscape Evolut Birmensdorf Switzerland Univ Grenoble Alpes Univ Savoie MontBlanc CNRS Lab Ecol Alpine Grenoble France Univ Montpellier MARBEC CNRS IFREMERIRD Montpellier France Inst Univ France Paris France SPYGEN Le Bourget Du Lac France

Environmental DNA (eDNA) metabarcoding provides an efficient approach for documenting biodiversity patterns in marine and terrestrial ecosystems. The complexity of these data prevents current methods from extracting and analyzing all the relevant ecological information they contain, and new methods may provide better dimensionality reduction and clustering. Here we present two new deep learning-based methods that combine different types of neural networks (NNs) to ordinate eDNA samples and visualize ecosystem properties in a two-dimensional space: the first is based on variational autoencoders and the second on deep metric learning. The strength of our new methods lies in the combination of two inputs: the number of sequences found for each molecular operational taxonomic unit (MOTU) detected and their corresponding nucleotide sequence. Using three different datasets, we show that our methods accurately represent several biodiversity indicators in a two-dimensional latent space: MOTU richness per sample, sequence a-diversity per sample, Jaccard's and sequence beta-diversity between samples. We show that our nonlinear methods are better at extracting features from eDNA datasets while avoiding the major biases associated with eDNA. Our methods outperform traditional dimension reduction methods such as Principal Component Analysis, t-distributed Stochastic Neighbour Embedding, Nonmetric Multidimensional Scaling and Uniform Manifold Approximation and Projection for dimension reduction. Our results suggest that NNs provide a more efficient way of extracting structure from eDNA metabarcoding data, thereby improving their ecological interpretation and thus biodiversity monitoring.

关键词： biodiversity monitoring data visualization deep learning deep metric learning environmental DNA machine learning neural networks variational autoencoder

来源：评论

学校读者我要写书评

暂无评论

Grasp-O: A Generative System for Object-Centric 6-DoF Grasping of Unknown Objects 15th

Grasp-O: A Generative System for Object-Centric 6-DoF Graspi...

引用

15th European Robotics Forum

作者： Barad, Kuldeep R. Orsulak, Andrej Richards, Antoine Dentler, Jan Olivares-Mendez, Miguel Martinec, Carol Univ Luxembourg Interdisciplinary Ctr Secur Reliabil & Trust SnT SpaceR L-1855 Esch Sur Alzette Luxembourg Redwire Space Europe L-2530 Luxembourg Luxembourg

ISBN: (纸本)9783031764301;9783031764288;9783031764271

Generative models are a promising avenue for learning generalizable robotic tasks from data. A fundamental task that remains a challenge to autonomous manipulation is the 6-DoF grasping of unknown objects. This work proposes Grasp-O: a simple, fast, and robust system for general-purpose vision-based 6-DoF grasping applications. Our system is built using a powerful and efficient variational autoencoder (VAE) that learns a distribution of SE(3) grasp poses conditioned on object point clouds. The generative model is complemented by a grasp classification network that discriminates between good and bad grasp. We conduct extensive evaluations in simulation and the real world and demonstrate that our system outperforms existing VAE-based methods.

关键词： vision-based grasping variational autoencoder point clouds

来源：评论

学校读者我要写书评

暂无评论

SIFISINGER: A HIGH-FIDELITY END-TO-END SINGING VOICE SYNTHESIZER BASED ON SOURCE-FILTER MODEL 49

SIFISINGER: A HIGH-FIDELITY END-TO-END SINGING VOICE SYNTHES...

引用

49th IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Cui, Jianwei Gu, Yu Weng, Chao Zhang, Jie Chen, Liping Dai, Lirong Univ Sci & Technol China USTC NERC SLIP Hefei Peoples R China Tencent AI Lab Shenzhen Peoples R China

ISBN: (纸本)9798350344868;9798350344851

This paper presents an advanced end-to-end singing voice synthesis (SVS) system based on the source-filter mechanism that directly translates lyrical and melodic cues into expressive and high-fidelity human-like singing. Similarly to VISinger 2, the proposed system also utilizes training paradigms evolved from VITS and incorporates elements like the fundamental pitch (F0) predictor and waveform generation decoder. To address the issue that the coupling of mel-spectrogram features with F0 information may introduce errors during F0 prediction, we consider two strategies. Firstly, we leverage mel-cepstrum (mcep) features to decouple the intertwined melspectrogram and F0 characteristics. Secondly, inspired by the neural source-filter models, we introduce source excitation signals as the representation of F0 in the SVS system, aiming to capture pitch nuances more accurately. Meanwhile, differentiable mcep and F0 losses are employed as the waveform decoder supervision to fortify the prediction accuracy of speech envelope and pitch in the generated speech. Experiments on the Opencpop dataset demonstrate efficacy of the proposed model in synthesis quality and intonation accuracy. Synthesized audio samples are available at: https://***/sifisinger.

关键词： Singing voice synthesis variational autoencoder adversarial learning neural source-filter model VITS

来源：评论

学校读者我要写书评

暂无评论

Deep Learning Enabled Bandpass Electromagnetic Metasurface Inverse Design

Deep Learning Enabled Bandpass Electromagnetic Metasurface I...

引用

2024 Cross Strait Radio Science and Wireless Technology Conference

作者： Zhao, Shenghui Zhang, Ruizhi Guo, Meng Zhang, Xuewei Nie, Lifeng Pang, Xiaoyu Res Inst Special Structures Aeronaut Composite AV Jinan 250023 Shandong Peoples R China Qilu Univ Technol Shandong Acad Sci Natl Supercomput Ctr JinanKey Lab Comp Power Net Shandong Comp Sci CtrMinist Educ Jinan 250013 Shandong Peoples R China

ISBN: (纸本)9798331507800;9798331507794

Conditional variational autoencoders (cVAE) is proposed to automate and optimize the design of band-pass electromagnetic metasurfaces. Electromagnetic responses curves and metasurface structural parameters are used as the input to train the deep neural work based cVAE model. The model is trained on ten thousand metasurface and transmission curve pairs, which are calculated using MIT Electromagnetic Equation Propagation (MEEP) software. The trained model takes transmission curves as input and then output metasurface structures, achieving a deep learning enabled inverse design process. By comparing the input transmission curve with the MEEP simulated curves of the generated metasurface, it is found that the model has high accuracy, giving a mean square error (MSE) value of 0.005 and R2 score of 0.967. The metasurface structures generated by cVAE also show good diversity, indicating that the cVAE captures the effective features in the latent space. Using the cVAE model, metasurface could be generated in less than one second using a single graphics processing unit (CPU), paving the way for fast, efficient and scalable inverse design solutions in the field of electromagnetic metasurfaces.

关键词： metasurface frequency selective surface deep learning variational autoencoder

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：