检索结果-内蒙古大学图书馆

Asia-Pacific-Signal-and-Information-Processing-Association Annual Summit and Conference (APSIPA ASC)

作者： Lo, Chun-Huang Lee, Chung-Nan Natl Sun Yat Sen Univ Dept Comp Sci & Engn Kaohsiung Taiwan

ISBN: (纸本)9798350300673

This paper presents a vision-based unmanned aerial vehicle (UAV) indoor obstacle avoidance using a deep reinforcement learning (DRL). The system consists of two parts a depth map compression and a UAV control. For the depth map compression part, the pre-trained variational autoencoder model is used to improve obstacle avoidance and reduce the training time. The states include the UAV information, gray images and compressed features. A dueling double deep recurrent Q network model is used to control the UAV. The network is trained in the AirSim simulation. To validate the performance of the proposed algorithm some simulations are conducted. The results show that the proposed algorithm can avoid obstacles at a fast speed in a narrow space, and fly through a difficult L-shaped corner in an indoor simulation.

关键词： Unmanned Aerial Vehicle Obstacle Avoidance Deep Reinforcement Learning AirSim variational autoencoder

来源：评论

学校读者我要写书评

暂无评论

Generative Slate Recommendation with Reinforcement Learning 23

Generative Slate Recommendation with Reinforcement Learning

引用

16th International Conference on Web Search and Data Mining

作者： Deffayet, Romain Thonet, Thibaut Renders, Jean-Michel de Rijke, Maarten Naver Labs Europe Meylan France Univ Amsterdam Amsterdam Netherlands

ISBN: (纸本)9781450394079

Recent research has employed reinforcement learning (RL) algorithms to optimize long-term user engagement in recommender systems, thereby avoiding common pitfalls such as user boredom and filter bubbles. They capture the sequential and interactive nature of recommendations, and thus offer a principled way to deal with long-term rewards and avoid myopic behaviors. However, RL approaches are intractable in the slate recommendation scenario - where a list of items is recommended at each interaction turn due to the combinatorial action space. In that setting, an action corresponds to a slate that may contain any combination of items. While previous work has proposed well-chosen decompositions of actions so as to ensure tractability, these rely on restrictive and sometimes unrealistic assumptions. Instead, in this work we propose to encode slates in a continuous, low-dimensional latent space learned by a variational auto-encoder. Then, the RL agent selects continuous actions in this latent space, which are ultimately decoded into the corresponding slates. By doing so, we are able to (i) relax assumptions required by previous work, and (ii) improve the quality of the action selection by modeling full slates instead of independent items, in particular by enabling diversity. Our experiments performed on a wide array of simulated environments confirm the effectiveness of our generative modeling of slates over baselines in practical scenarios where the restrictive assumptions underlying the baselines are lifted. Our findings suggest that representation learning using generative models is a promising direction towards generalizable RL-based slate recommendation.

关键词： Slate recommendation Reinforcement learning variational autoencoder

来源：评论

学校读者我要写书评

暂无评论

Two-stage instrument timbre transfer method using RAVE 26

Two-stage instrument timbre transfer method using RAVE

引用

26th International Symposium on Multimedia, ISM 2024

作者： Hu, Di Ito, Katunobu Hosei University Graduate School of Computer and Information Sciences Tokyo Japan Hosei University Faculty of Computer and Information Sciences Tokyo Japan

ISBN: (纸本)9798331511111

Recently, the real-time audio variational autoencoder (RAVE) method was developed for high-quality audio waveform synthesis. The RAVE method is based on a variational autoencoder and employs a two-stage training strategy. However, the RAVE model still has limitations in timbre transformation, especially when converting between instruments with significantly different timbres. Issues such as pitch instability, inaccurate timbre reproduction, and severe degradation in sound quality can arise. To enhance timbre transfer performance, we propose a two-stage timbre transformation method using RAVE, which involves applying two timbre transfer models to perform a dual transformation on the original input audio. To evaluate the proposed method, we trained the model and tested its performance using audio generated from MIDI and SoundFont2 sound sources. The results demonstrate that the proposed method improves timbre transfer compared to the single-stage RAVE model. © 2024 IEEE.

关键词： audio synthesis timbre transfer variational autoencoder

来源：评论

学校读者我要写书评

暂无评论

Data-driven fault diagnosis based on the integrated deep nonlinear dynamic system model 43

Data-driven fault diagnosis based on the integrated deep non...

引用

43rd Chinese Control Conference, CCC 2024

作者： Tang, Xiaochu Tao, Na Zhang, Yi Li, Yuan Shenyang Aerospace University School of Automation Shenyang China Shenyang University of Chemical Technology College of Information Engineering Shenyang China

ISBN: (纸本)9789887581581

To ensure the safety and reliability of complex industrial processes are very important. Therefore, extracting multiple features of data effectively is a great significance to improve the accuracy of modeling for fault diagnosis. Dynamic, uncertainty and nonlinearity are main characteristics of industrial process data. However, it is challenging for modeling to extract multiple features of process data at the same time. In this paper, an integrated nonlinear dynamic system model is proposed for fault diagnosis based on variational autoencoder-linear dynamic system (VAE-LDS). First, the deep learning algorithm variational autoencoder (VAE) is used to extract the nonlinear data feature and learn the potential representation of data. Furthermore, the VAE model is embedded into the linear dynamic system (LDS) so that the dynamics and uncertainty underlying data can be extracted simultaneously. In this way, A comprehensive model integrating multiple features can be established. Finally, the proposed method is applied to the TE process for fault diagnosis comparing with other methods. The results show the proposed method has superior performance. © 2024 Technical Committee on Control Theory, Chinese Association of Automation.

关键词： Data-driven Fault Detection Fault Diagnosis Linear Dynamic System variational autoencoder

来源：评论

学校读者我要写书评

暂无评论

An Ultrasound-Based Surveillance System for Bathroom Posture and Location Estimation

An Ultrasound-Based Surveillance System for Bathroom Posture...

引用

2024 IEEE International Conference on Consumer Electronics, ICCE 2024

作者： Sato, Shun Ohara, Ryotaro Kamarulzaman, M. Shahrul Amir Yasuda, Yuto Izumi, Shintaro Kawaguchi, Hiroshi Kobe University School of System Informatics Kobe Japan Kobe University School of Science Technology and Innovation Kobe Japan

ISBN: (纸本)9798350324136

Bathrooms can be slippery, increasing the risk of falling. In addition, because people enter the bathroom alone, it is difficult to detect accidents immediately when they occur. Therefore, a system is required to quickly detect accidents and call for assistance. In this study, the location and the posture of the bather are important attributes. We propose a method to estimate the location of a bather using ultrasound correlation values as input to a deep neural network and to estimate the bather's posture. Therefore, the mean absolute error achieved 11.7 cm for location estimation, and accuracy achieved 95.3% for posture estimation in the experiment involving an individual moving throughout the bathroom. © 2024 IEEE.

关键词： Imaging Location Estimation Posture Estimation Ultrasound variational autoencoder

来源：评论

学校读者我要写书评

暂无评论

Hierarchical and Self-Attended Sequence autoencoder

引用

IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 2022年第9期44卷 4975-4986页

作者： Chien, Jen-Tzung Wang, Chun-Wei Natl Chiao Tung Univ Dept Elect & Comp Engn Hsinchu 30010 Taiwan

It is important and challenging to infer stochastic latent semantics for natural language applications. The difficulty in stochastic sequential learning is caused by the posterior collapse in variational inference. The input sequence is disregarded in the estimated latent variables. This paper proposes three components to tackle this difficulty and build the variational sequence autoencoder (VSAE) where sufficient latent information is learned for sophisticated sequence representation. First, the complementary encoders based on a long short-term memory (LSTM) and a pyramid bidirectional LSTM are merged to characterize global and structural dependencies of an input sequence, respectively. Second, a stochastic self attention mechanism is incorporated in a recurrent decoder. The latent information is attended to encourage the interaction between inference and generation in an encoder-decoder training procedure. Third, an autoregressive Gaussian prior of latent variable is used to preserve the information bound. Different variants of VSAE are proposed to mitigate the posterior collapse in sequence modeling. A series of experiments are conducted to demonstrate that the proposed individual and hybrid sequence autoencoders substantially improve the performance for variational sequential learning in language modeling and semantic understanding for document classification and summarization.

关键词： Decoding Stochastic processes Training Semantics Recurrent neural networks Natural languages Data models Sequence generation recurrent neural network variational autoencoder hierarchical model self attention

来源：评论

学校读者我要写书评

暂无评论

Enhancing variational Generation Through Self-Decomposition

引用

IEEE ACCESS 2022年 10卷 67510-67520页

作者： Asperti, Andrea Bugo, Laura Filippini, Daniele Univ Bologna Dept Informat Sci & Engn DISI I-40126 Bologna Italy

In this article we introduce the notion of Split variational autoencoder (SVAE), whose output (x) over cap is obtained as a weighted sum sigma circle dot (x) over cap (1) + (1 - sigma) circle dot (x) over cap (2) of two generated images (x) over cap (1);(x) over cap (2), and sigma is a learned compositional map. The composing images (x) over cap (1);(x) over cap (2), as well as the sigma-map are automatically synthesized by the model. The network is trained as a usual variational autoencoder with a negative loglikelihood loss between training and reconstructed images. No additional loss is required for (x) over cap (1);(x) over cap (2) or sigma, neither any form of human tuning. The decomposition is nondeterministic, but follows two main schemes, that we may roughly categorize as either "syntactic'' or "semantic.'' In the first case, the map tends to exploit the strong correlation between adjacent pixels, splitting the image in two complementary high frequency sub-images. In the second case, the map typically focuses on the contours of objects, splitting the image in interesting variations of its content, with more marked and distinctive features. In this case, according to empirical observations, the Frechet Inception Distance (FID) of (x) over cap (1) and (x) over cap (2) is usually lower (hence better) than that of Ox, that clearly suffers from being the average of the former. In a sense, a SVAE forces the variational autoencoder to make choices, in contrast with its intrinsic tendency to average between alternatives with the aim to minimize the reconstruction loss towards a specific sample. According to the FID metric, our technique, tested on typical datasets such as Mnist, Cifar10 and CelebA, allows us to outperform all previous purely variational architectures (not relying on normalization flows).

关键词： Deep learning generative modeling multi-layer neural networks representation learning unsupervised learning variational autoencoder

来源：评论

学校读者我要写书评

暂无评论

Reinforcement Learning With Vision-Proprioception Model for Robot Planar Pushing

引用

FRONTIERS IN NEUROROBOTICS 2022年第0期16卷 829437页

作者： Cong, Lin Liang, Hongzhuo Ruppel, Philipp Shi, Yunlei Goerner, Michael Hendrich, Norman Zhang, Jianwei Univ Hamburg Dept Informat TAMS Grp Hamburg Germany

We propose a vision-proprioception model for planar object pushing, efficiently integrating all necessary information from the environment. A variational autoencoder (VAE) is used to extract compact representations from the task-relevant part of the image. With the real-time robot state obtained easily from the hardware system, we fuse the latent representations from the VAE and the robot end-effector position together as the state of a Markov Decision Process. We use Soft Actor-Critic to train the robot to push different objects from random initial poses to target positions in simulation. Hindsight Experience replay is applied during the training process to improve the sample efficiency. Experiments demonstrate that our algorithm achieves a pushing performance superior to a state-based baseline model that cannot be generalized to a different object and outperforms state-of-the-art policies which operate on raw image observations. At last, we verify that our trained model has a good generalization ability to unseen objects in the real world.

关键词： reinforcement learning robot manipulation planar pushing multimodal variational autoencoder Soft Actor-Critic

来源：评论

学校读者我要写书评

暂无评论

Generative Feature Extraction From Sentinel 1 and 2 Data for Prediction of Forest Aboveground Biomass in the Italian Alps

引用

IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING 2022年 15卷 4755-4771页

作者： Naik, Parth Dalponte, Michele Bruzzone, Lorenzo Univ Trento Dept Informat Engn & Comp Sci I-38123 Trento Italy Fdn Edmund Mach Res & Innovat Ctr I-38098 San Michele All Adige Italy

Aboveground biomass (AGB) is an important forest attribute directly linked to the forest carbon pool. The use of satellite remote sensing (RS) data has increased for AGB prediction due to their large footprint and low-cost availability. However, they have been limited due to saturation effect that leads to low prediction precision. In this article, we propose an innovative and dynamic architecture based on generative neural network that extracts target oriented generative features for predicting forest AGB using satellite RS data. These features are more resilient to mixed forest types and geographical conditions as compared to the traditional features and models. The effectiveness of the proposed features was assessed by experiments performed using multispectral, synthetic aperture radar, and combined dual-source datasets. The proposed model achieved best performance in terms of precision, model agreement, and overfitting as compared to the other conventional models for all analyzed datasets. The t-distributed stochastic neighbor embedding scatterplots of the generative features clearly show one dimension of the feature space associated with the target AGB. Feature importance analysis indicated that the produced generative features were more significant than the conventional analytical features. Also, the model provided a robust framework for homogeneous fusion of multisensor features from satellite RS data for predicting AGB.

关键词： Forestry Feature extraction Synthetic aperture radar Data models Vegetation Satellite broadcasting Mathematical models Aboveground biomass (AGB) feature extraction feature fusion generative features variational autoencoder

来源：评论

学校读者我要写书评

暂无评论

Investigating Deep Learning Based Breast Cancer Subtyping Using Pan-Cancer and Multi-Omic Data

引用

IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS 2022年第1期19卷 121-134页

作者： Cristovao, Francisco Cascianelli, Silvia Canakoglu, Arif Carman, Mark Nanni, Luca Pinoli, Pietro Masseroli, Marco Politecn Milan Dipartimento Elettron Informaz & Bioingn I-20133 Milan Italy

Breast Cancer comprises multiple subtypes implicated in prognosis. Existing stratification methods rely on the expression quantification of small gene sets. Next Generation Sequencing promises large amounts of omic data in the next years. In this scenario, we explore the potential of machine learning and, particularly, deep learning for breast cancer subtyping. Due to the paucity of publicly available data, we leverage on pan-cancer and non-cancer data to design semi-supervised settings. We make use of multi-omic data, including microRNA expressions and copy number alterations, and we provide an in-depth investigation of several supervised and semi-supervised architectures. Obtained accuracy results show simpler models to perform at least as well as the deep semi-supervised approaches on our task over gene expression data. When multi-omic data types are combined together, performance of deep models shows little (if any) improvement in accuracy, indicating the need for further analysis on larger datasets of multi-omic data as and when they become available. From a biological perspective, our linear model mostly confirms known gene-subtype annotations. Conversely, deep approaches model non-linear relationships, which is reflected in a more varied and still unexplored set of representative omic features that may prove useful for breast cancer subtyping.

关键词： Deep learning genomics multi-omics semi supervised 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：