Spatio-Temporal Multi-Task Learning via Tensor Decomposition

Xu, Jianpeng; Zhou, Jiayu; Tan, Pang‐Ning; Liu, Xi; Liu, Luo

doi:10.1109/tkde.2019.2956713

Cited by 15 publications

(10 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The resulting algorithm is named multiway multiblock covariates regression models. Another example for ensemble learning for the two-stage framework is presented in Xu et al (2019)…”

Section: Tensor Principal Component Regressionmentioning

confidence: 99%

“…The version of record is available at: http://dx.doi.org/10.1561/2200000087 by CP decomposition (Xu et al, 2019)…”

Section: It First Decomposes the Predictor Tensormentioning

confidence: 99%

“…The most common strategy for achieving online learning is to fit fresh data samples while maintaining the estimated model as close to the prior model as possible. Mathematically, the model for online learning is formulated in Xu et al (2019) as…”

Section: Online Learningmentioning

confidence: 99%

“…The collection of model parameters and the solution procedure for different regression models can differ. As an incremental learning algorithm for the regression model (4.103), WISDOM (Xu et al, 2019) jointly minimizes the objective function in model (4.103) for the newly arrived dataset and the regularization term (6.6) where {W, V, A, B, C} are old model parameters as defined in (4.103), { W, Ṽ, Ã, B, C} are new parameters that we need to estimate when a new dataset arrives. The estimated new parameters can be made similar to the prior ones by minimizing H(T , T ).…”

Section: Online Learningmentioning

confidence: 99%

See 3 more Smart Citations

Tensor Regression

Liu

Zhu

Zhang

et al. 2021

FNT in Machine Learning

View full text Add to dashboard Cite

The presence of multidirectional correlations in emerging multidimensional data poses a challenge to traditional regression modeling methods. Traditional modeling methods based on matrix or vector, for example, not only overlook the data's multidimensional information and lower model performance, but also add additional computations and storage requirements. Driven by the recent advances in applied mathematics, tensor regression has been widely used and proven effective in many fields, such as sociology, climatology, geography, economics, computer vision, chemometrics, and neuroscience. Tensor regression can explore multidirectional relatedness, reduce the number of model parameters and improve model robustness and efficiency. It is timely and valuable to summarize the developments of tensor regression in recent years and discuss promising future directions, which will help accelerate the research process of tensor regression, broaden the research direction, and provide tutorials for researchers interested in high dimensional regression tasks.

show abstract

“…The resulting algorithm is named multiway multiblock covariates regression models. Another example for ensemble learning for the two-stage framework is presented in Xu et al (2019)…”

Section: Tensor Principal Component Regressionmentioning

confidence: 99%

“…The version of record is available at: http://dx.doi.org/10.1561/2200000087 by CP decomposition (Xu et al, 2019)…”

Section: It First Decomposes the Predictor Tensormentioning

confidence: 99%

Section: Online Learningmentioning

confidence: 99%

Section: Online Learningmentioning

confidence: 99%

See 2 more Smart Citations

Tensor Regression

Liu

Zhu

Zhang

et al. 2021

FNT in Machine Learning

View full text Add to dashboard Cite

show abstract

“…Specifically in the ML community, Tensor Factorization (TF), as a classic dimensionality reduction technique, plays a key role for low-rank representation. Xu et al, [13] proposed a Spatio-temporal multi-task learning model via TF and in this work, tensor data is of 5-order, i.e., weather, traffic volume, crime rate, disease incidents, and time. Meanwhile, this model made predictions through the time-order for the multi-task in weather, traffic volume, crime rate, and disease incidents orders and the relationship construction between those orders is via TF.…”

mentioning

confidence: 99%

SGD_Tucker: A Novel Stochastic Optimization Strategy for Scalable Parallel Sparse Tucker Decomposition

Rellermeyer

et al. 2021

IEEE Trans. Parallel Distrib. Syst.

View full text Add to dashboard Cite

Sparse Tucker Decomposition (STD) algorithms learn a core tensor and a group of factor matrices to obtain an optimal low-rank representation feature for the High-Order, High-Dimension, and Sparse Tensor (HOHDST). However, existing STD algorithms face the problem of intermediate variables explosion which results from the fact that the formation of those variables, i.e., matrices Khatri-Rao product, Kronecker product, and matrix-matrix multiplication, follows the whole elements in sparse tensor. The above problems prevent deep fusion of efficient computation and big data platforms. To overcome the bottleneck, a novel stochastic optimization strategy (SGD Tucker) is proposed for STD which can automatically divide the high-dimension intermediate variables into small batches of intermediate matrices. Specifically, SGD Tucker only follows the randomly selected small samples rather than the whole elements, while maintaining the overall accuracy and convergence rate. In practice, SGD Tucker features the two distinct advancements over the state of the art. First, SGD Tucker can prune the communication overhead for the core tensor in distributed settings. Second, the low data-dependence of SGD Tucker enables fine-grained parallelization, which makes SGD Tucker obtaining lower computational overheads with the same accuracy. Experimental results show that SGD Tucker runs at least 2X faster than the state of the art.

show abstract

Tensor Regression

Liu

Zhang

et al. 2021

Tensor Computation for Data Analysis

View full text Add to dashboard Cite

Spatio-Temporal Multi-Task Learning via Tensor Decomposition

Cited by 15 publications

References 35 publications

Tensor Regression

Tensor Regression

SGD_Tucker: A Novel Stochastic Optimization Strategy for Scalable Parallel Sparse Tucker Decomposition

Tensor Regression

Contact Info

Product

Resources

About