检索结果-内蒙古大学图书馆

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

作者： Zihan Wang Yunhang Shen Mengtian Li Ke Li Xing Sun Shaohui Lin Lizhuang Ma East China Normal University Tencent Youtu Lab Shanghai University Key Laboratory of Advanced Theory and Application in Statistics and Data Science

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Current weakly supervised point cloud semantic segmentation struggles with insufficient utilization of limited annotations in unimodal representation learning due to the sparse and textureless nature of point clouds. In this work, we leverage cross-modality information by transferring knowledge from image and text sources to the point cloud network. The intuition is that images contribute rich texture, color, and discriminative information, complementing point clouds to boost semantic segmentation performance. To reduce extensive computational resources for cross-modality fusion, we introduce the Multi-Scale Deformable Knowledge Transfer, an innovative training scheme that optimizes and extends the one-to-one mapping to flexible one-to-many relations between multi-modal data. Furthermore, we employ pre-trained image-text models to generate pseudo labels for point clouds and construct positive and negative samples for semantic contrastive regularization, facilitating the full exploitation of unlabeled data. The experimental results evaluated on SemanticKITTI and nuScenes demonstrate substantial improvements, achieving an average gain of 3.8% over the previous weakly supervised methods, and comparable performances to fully supervised approaches.

关键词： Point cloud compression Training Representation learning Weak supervision Three-dimensional displays Semantic segmentation Semantics Signal processing Speech processing Knowledge transfer

来源：评论

学校读者我要写书评

暂无评论

Estimation of Scale Transformation for Approximate Periodic Time Series with Long-Term Trend

引用

Journal of Mathematical Research with applications 2021年第3期41卷 238-258页

作者： Shujin WU Key Laboratory of Advanced Theory and Application in Statistics and Data Science-MOE School of Statistics Faculty of Economics and Management East China Normal University

Approximate periodic time series means it has an approximate periodic trend. The so-called approximate periodicity refers that it looks like having periodicity, however the length of each period is not constant such as sunspot data. Approximate periodic time series has a wide application prospect in modelling social economic phenomenon. As for approximate periodic time series, the key problem is to depict its approximate periodic trend because it can be dealt as an ordinary time series only if its approximate periodic trend has been depicted. However,there is little study on depicting approximate periodic *** the paper, the authors first establish some necessary theories, especially bring forward the concept of shape-retention transformation with lengthwise compression and obtain necessary and sufficient condition for linear shape-retention transformation with lengthwise compression,then basing on the theories the authors present a method to estimate scale transformation, which can model approximate periodic trend very clearly. At last, a simulated example is analyzed by this presented method. The results show that the presented method is very effective and very powerful.

关键词： time series approximate periodicity scale transformation shape-retention transformation with lengthwise compression

来源：评论

学校读者我要写书评

暂无评论

Dynamic Contrastive Knowledge Distillation for Efficient Image Restoration 39

Dynamic Contrastive Knowledge Distillation for Efficient Ima...

引用

39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

作者： Zhou, Yunshuai Qiao, Junbo Liao, Jincheng Li, Wei Li, Simiao Xie, Jiao Shen, Yunhang Hu, Jie Lin, Shaohui East China Normal University Shanghai China Huawei Noah's Ark Lab China Xiamen University China Key Laboratory of Advanced Theory and Application in Statistics and Data Science - MOE China

ISBN: (纸本)157735897X

Knowledge distillation (KD) is a valuable yet challenging approach that enhances a compact student network by learning from a high-performance but cumbersome teacher model. However, previous KD methods for image restoration overlook the state of the student during the distillation, adopting a fixed solution space that limits the capability of KD. Additionally, relying solely on L1-type loss struggles to leverage the distribution information of images. In this work, we propose a novel dynamic contrastive knowledge distillation (DCKD) framework for image restoration. Specifically, we introduce dynamic contrastive regularization to perceive the student's learning state and dynamically adjust the distilled solution space using contrastive learning. Additionally, we also propose a distribution mapping module to extract and align the pixel-level category distribution of the teacher and student models. Note that the proposed DCKD is a structure-agnostic distillation framework, which can adapt to different backbones and can be combined with methods that optimize upper-bound constraints to further enhance model performance. Extensive experiments demonstrate that DCKD significantly outperforms the state-of-the-art KD methods across various image restoration tasks and backbones. Copyright © 2025, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

A hybrid deep learning method for controlled stochastic Kolmogorov systems with regime-switching

A hybrid deep learning method for controlled stochastic Kolm...

引用

International Conference on Control, Decision and Information Technologies (CoDIT)

作者： Yu Zhang Zhuo Jin Jiaqin Wei School of Statistics Key Laboratory of Advanced Theory and Application in Statistics and Data Science-MOE East China Normal University Shanghai China Department of Actuarial Studies and Business Analytics Macquarie University NSW Australia

ISBN: (数字)9798350373974

ISBN: (纸本)9798350373981

In this paper, we employ numerical methods based on deep learning algorithms for solving controlled stochastic Kolmogorov systems with regime-switching. Different from classical control problems, each component of the state in controlled Kolmogorov systems is nonnegative. Due to the nonlinearity and complexity of the controlled stochastic Kolmogorov systems, we develop a hybrid deep learning method to numerically solve the optimal controls under this system. Subsequently, we apply the hybrid deep learning method to solve a specific case of a controlled stochastic Kolmogorov system, specifically controlled SIS (susceptible-infected-susceptible) systems. Finally, the effectiveness of the proposed hybrid deep learning method is verified through numerical results.

关键词： Deep learning Closed-form solutions Neural networks Optimal control Aerospace electronics Control systems Complexity theory Approximation methods Information technology Method of moments

来源：评论

学校读者我要写书评

暂无评论

Estimation and inference of high-dimensional partially linear regression models with latent factors

arXiv

引用

arXiv 2025年

作者： Shi, Yanmei Hao, Meiling Tang, Yanlin Guo, Xu School of Statistics Beijing Normal University Beijing China School of Statistics University of International Business and Economics Beijing China Key Laboratory of Advanced Theory and Application in Statistics and Data Science–MOE School of Statistics East China Normal University Shanghai China

In this paper, we introduce a novel high-dimensional Factor-Adjusted sparse Partially Linear regression Model (FAPLM), to integrate the linear effects of high-dimensional latent factors with the nonparametric effects of low-dimensional covariates. The proposed FAPLM combines the interpretability of linear models, the flexibility of nonparametric models, with the ability to effectively capture the dependencies among high-dimensional covariates. We develop a penalized estimation approach for the model by leveraging B-spline approximations and factor analysis techniques. Theoretical results establish error bounds for the estimators, aligning with the minimax rates of standard Lasso problems. To assess the significance of the linear component, we introduce a factor-adjusted projection debiased procedure and employ the Gaussian multiplier bootstrap method to derive critical values. Theoretical guarantees are provided under regularity conditions. Comprehensive numerical experiments validate the finite-sample performance of the proposed method. Its successful application to a birth weight dataset, the motivating example for this study, highlights both its effectiveness and practical relevance. Copyright © 2025, The Authors. All rights reserved.

关键词： Factor analysis

来源：评论

学校读者我要写书评

暂无评论

DYNAMIC-LLAVA: EFFICIENT MULTIMODAL LARGE LANGUAGE MODELS VIA DYNAMIC VISION-LANGUAGE CONTEXT SPARSIFICATION

arXiv

引用

arXiv 2024年

作者： Huang, Wenxuan Zhai, Zijie Shen, Yunhang Cao, Shaosheng Zhao, Fei Xu, Xiangfeng Ye, Zheyu Hu, Yao Lin, Shaohui East China Normal University China Xiamen University China Xiaohongshu Inc China Nanjing University China Key Laboratory of Advanced Theory and Application in Statistics and Data Science MOE China

Multimodal Large Language Models (MLLMs) have achieved remarkable success in vision understanding, reasoning, and interaction. However, the inference computation and memory increase progressively with the generation of output tokens during decoding, directly affecting the efficacy of MLLMs. Existing methods attempt to reduce the vision context redundancy to achieve efficient MLLMs. Unfortunately, the efficiency benefits of the vision context reduction in the prefill stage gradually diminish during the decoding stage. To address this problem, we proposed a dynamic vision-language context sparsification framework Dynamic-LLaVA, which dynamically reduces the redundancy of vision context in the prefill stage and decreases the memory and computation overhead of the generated language context during decoding. Dynamic-LLaVA designs a tailored sparsification inference scheme for different inference modes, i.e., prefill, decoding with and without KV cache, to achieve efficient inference of MLLMs. In practice, Dynamic-LLaVA can reduce computation consumption by ∼75% in the prefill stage. Meanwhile, throughout the entire generation process of MLLMs, Dynamic-LLaVA reduces the ∼50% computation consumption under decoding without KV cache, while saving ∼50% GPU memory overhead when decoding with KV cache, due to the vision-language context sparsification. Extensive experiments also demonstrate that Dynamic-LLaVA achieves efficient inference for MLLMs with negligible understanding and generation ability degradation or even performance gains compared to the full-context inference baselines. Code is available at https://***/Osilly/dynamic_llava . Copyright © 2024, The Authors. All rights reserved.

关键词： Decoding

来源：评论

学校读者我要写书评

暂无评论

Dynamic Contrastive Knowledge Distillation for Efficient Image Restoration

arXiv

引用

arXiv 2024年

作者： Zhou, Yunshuai Qiao, Junbo Liao, Jincheng Li, Wei Li, Simiao Xie, Jiao Shen, Yunhang Hu, Jie Lin, Shaohui East China Normal University Shanghai China Huawei Noah’s Ark Lab China Xiamen University China Key Laboratory of Advanced Theory and Application in Statistics and Data Science - MOE China

Knowledge distillation (KD) is a valuable yet challenging approach that enhances a compact student network by learning from a high-performance but cumbersome teacher model. However, previous KD methods for image restoration overlook the state of the student during the distillation, adopting a fixed solution space that limits the capability of KD. Additionally, relying solely on L1-type loss struggles to leverage the distribution information of images. In this work, we propose a novel dynamic contrastive knowledge distillation (DCKD) framework for image restoration. Specifically, we introduce dynamic contrastive regularization to perceive the student’s learning state and dynamically adjust the distilled solution space using contrastive learning. Additionally, we also propose a distribution mapping module to extract and align the pixel-level category distribution of the teacher and student models. Note that the proposed DCKD is a structure-agnostic distillation framework, which can adapt to different backbones and can be combined with methods that optimize upper-bound constraints to further enhance model performance. Extensive experiments demonstrate that DCKD significantly outperforms the state-of-the-art KD methods across various image restoration tasks and backbones. Our codes are available at https://***/super-SSS/DCKD. Copyright © 2024, The Authors. All rights reserved.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Semi-supervised inference for block-wise missing data without imputation

The Journal of Machine Learning Research

引用

The Journal of Machine Learning Research 2024年第1期25卷 4902-4937页

作者： Shanshan Song Yuanyuan Lin Yong Zhou School of Mathematical Sciences and School of Economics and Management Tongji University Shanghai China Department of Statistics The Chinese University of Hong Kong Hong Kong China Key Laboratory of Advanced Theory and Application in Statistics and Data Science MOE Academy of Statistics and Interdisciplinary Sciences and School of Statistics East China Normal University Shanghai China

We consider statistical inference for single or low-dimensional parameters in a high-dimensional linear model under a semi-supervised setting, wherein the data are a combination of a labelled block-wise missing data set of a relatively small size and a large unlabelled data set. The proposed method utilises both labelled and unlabelled data without any imputation or removal of the missing observations. The asymptotic properties of the estimator are established under regularity conditions. Hypothesis testing for low-dimensional coefficients are also studied. Extensive simulations are conducted to examine the theoretical results. The method is evaluated on the Alzheimer's Disease Neuroimaging Initiative data.

关键词： block-missing data confidence intervals hypothesis testing semi-supervised inference

来源：评论

学校读者我要写书评

暂无评论

A review of distributed statistical inference

引用

Statistical theory and Related Fields 2022年第2期6卷 89-99页

作者： Yuan Gao Weidong Liu Hansheng Wang Xiaozhou Wang Yibo Yan Riquan Zhang School of Statistics and Key Laboratory of Advanced Theory and Application in Statistics and Data Science–MOE East China Normal UniversityShanghaiPeople’s Republic of China School of Mathematical Sciences and Key Lab of Articial Intelligence–MOE Shanghai Jiao Tong UniversityShanghaiPeople’s Republic of China Guanghua School of Management Peking UniversityBeijingPeople’s Republic of China

The rapid emergence of massive datasets in various fields poses a serious challenge to tra-ditional statistical ***,it provides opportunities for researchers to develop novel *** by the idea of divide-and-conquer,various distributed frameworks for statistical estimation and inference have been *** were developed to deal with large-scale statistical optimization *** paper aims to provide a comprehensive review for related *** includes parametric models,nonparametric models,and other frequently used *** key ideas and theoretical properties are *** trade-off between communication cost and estimate precision together with other concerns is discussed.

关键词： Distributed computing divide-and-conquer communication-efficiency shrinkage methods nonparametric estimation principal component analysis feature screening bootstrap

来源：评论

学校读者我要写书评

暂无评论

Partially Linear Single-Index Model in the Presence of Measurement Error

引用

Journal of Systems science & Complexity 2022年第6期35卷 2361-2380页

作者： LIN Hongmei SHI Jianhong TONG Tiejun ZHANG Riquan School of Statistics and Information Shanghai University of International Business and EconomicsShanghai 201620China Key Laboratory of Advanced Theory and Application in Statistics and Data Science Ministry of EducationEast China Normal UniversityShanghai 200062China School of Mathematics and Computer Science Shanxi Normal UniversityLinfen 041081China Department of Mathematics Hong Kong Baptist UniversityHongKong 519087China School of Statistics East China Normal UniversityShanghai 200062China

The partially linear single-index model(PLSIM) is a flexible and powerful model for analyzing the relationship between the response and the multivariate covariates. This paper considers the PLSIM with measurement error possibly in all the variables. The authors propose a new efficient estimation procedure based on the local linear smoothing and the simulation-extrapolation method,and further establish the asymptotic normality of the proposed estimators for both the index parameter and nonparametric link function. The authors also carry out extensive Monte Carlo simulation studies to evaluate the finite sample performance of the new method, and apply it to analyze the osteoporosis prevention data.

关键词： Local linear regression measurement error partially linear model SIMEX single-index model

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：