Tensors provide a structured representation for multidimensional data, yet discretization discards the underlying continuous structure when the data originate from continuous processes. We address this limitation by introducing a functional Tucker decomposition (FTD) that embeds a mode-wise continuity constraint directly into the factorization. The FTD models the continuous mode as a function in a reproducing kernel Hilbert space (RKHS), avoiding a prespecified basis while preserving the multilinear subspace structure of the Tucker model. We derive a reconstruction error bound for the continuous mode that quantifies the approximation quality when a subspace estimated on one domain is reused on another. This bound provides theoretical justification for subspace transfer, whose practical value we demonstrate on cross-domain classification tasks in hyperspectral imaging and multivariate time-series analysis.
HOTCAKE: Higher Order Tucker Articulated Kernels for Deeper CNN Compression
Institute for Clarity in Documentation(清晰度文献研究所);Inria Paris-Rocquencourt(法国国家信息与自动化研究所巴黎-罗康库尔中心);Rajiv Gandhi University(拉吉夫·甘地大学);Tsinghua University(清华大学);Palmer Research Laboratories(帕尔默研究实验室)
Conditional Hierarchical Bayesian Tucker Decomposition for Genetic Data Analysis
Engineering Sciences and Applied Mathematics(应用数学与工程科学);Industrial Engineering and Management Sciences(工业工程与管理科学);Preventive Medicine (Health and Biomedical Informatics)(预防医学(健康与生物医学信息学))
Efficient Nonnegative Tucker Decompositions: Algorithms and Uniqueness
Guangdong University of Technology(广东工业大学);RIKEN Brain Science Institute(理化学研究所脑科学研究所);Systems Research Institute, Polish Academy of Science(波兰科学院系统研究所)
DinTucker: Scaling up Gaussian process models on multidimensional arrays with billions of elements
Purdue University(普渡大学);IBM T. J. Watson Research Center(IBM T·J·沃森研究中心)
无限Tucker分解(InfTucker)和随机函数先验模型作为无限可交换阵列上的非参数贝叶斯模型,是比广泛使用的多线性分解方法(包括Tucker和PARAFAC分解)更强大的模型,(部分)由于它们能够建模阵列元素之间的非线性关系。尽管它们具有很好的预测性能和良好的理论基础,但由于训练时间过长,它们无法处理大量数据。为了克服这一限制,我们提出了分布式无限塔克(DINTUCKER),一个大规模的非线性张量分解算法的MAPREDUCE。在保持InfTucker的预测准确性的同时,它可以在海量数据上进行扩展。DINTUCKER基于一种新的分层贝叶斯模型,可以在子阵列上对InfTucker进行本地训练,并从所有本地训练结果中集成信息。我们使用分布式随机梯度下降,再加上变分推理,训练这个模型。我们将DINTUCKER应用于来自“Read the Web”项目(Carlson et al.,2010)和信息安全领域,并将其与最先进的大规模张量分解方法GigaTensor进行比较。在这两个数据集上,DINTUCKER以更少的计算时间实现了更高的预测精度。
Subspace Determination through Local Intrinsic Dimensional Decomposition: Theory and Experimentation
Gran Sasso Science Institute(格兰萨索科学研究所);Ecole Polytechnique de Montréal(蒙特利尔理工学院);National Institute of Informatics(国立信息学研究所);Saarland University(萨尔大学);University of Southern Denmark(南丹麦大学)