Abstract:CANDECOMP/PARAFAC (CP) tensor factorization of incomplete data is a powerful technique for tensor completion through explicitly capturing the multilinear latent factors. The existing CP algorithms require the tensor rank to be manually specified, however, the determination of tensor rank remains a challenging problem especially for CP rank. In addition, existing approaches do not take into account uncertainty information of latent factors, as well as missing entries. To address these issues, we formulate CP factorization using a hierarchical probabilistic model and employ a fully Bayesian treatment by incorporating a sparsity-inducing prior over multiple latent factors and the appropriate hyperpriors over all hyperparameters, resulting in automatic rank determination. To learn the model, we develop an efficient deterministic Bayesian inference algorithm, which scales linearly with data size. Our method is characterized as a tuning parameter-free approach, which can effectively infer underlying multilinear factors with a low-rank constraint, while also providing predictive distributions over missing entries. Extensive simulations on synthetic data illustrate the intrinsic capability of our method to recover the ground-truth of CP rank and prevent the overfitting problem, even when a large amount of entries are missing. Moreover, the results from real-world applications, including image inpainting and facial image synthesis, demonstrate that our method outperforms state-of-the-art approaches for both tensor factorization and tensor completion in terms of predictive performance.

Modelling Matrix Time Series Via a Tensor CP-decomposition

Identification and estimation for matrix time series CP-factor models

Estimation and Inference for CP Tensor Factor Models

High-Dimensional Vector Autoregressive Time Series Modeling via Tensor Decomposition

Contextual Tensor Decomposition by Projected Alternating Least Squares

Tensor Factor Model Estimation by Iterative Projection

CP Degeneracy in Tensor Regression

Tensor Networks for Latent Variable Analysis: Higher Order Canonical Polyadic Decomposition

Simultaneous Decorrelation of Matrix Time Series

Application Performance Modeling via Tensor Completion

Bayesian CP Factorization of Incomplete Tensors with Automatic Rank Determination

Canonical Polyadic Decomposition via the generalized Schur decomposition

Segmenting High-dimensional Matrix-valued Time Series via Sequential Transformations

Tensor rank learning in CP decomposition via convolutional neural network

NeuLFT: A Novel Approach to Nonlinear Canonical Polyadic Decomposition on High-Dimensional Incomplete Tensors

Inertial Accelerated Stochastic Mirror Descent for Large-Scale Generalized Tensor CP Decomposition

Bayesian inference of vector autoregressions with tensor decompositions

Tensor decompositions for count data that leverage stochastic and deterministic optimization

Probabilistic Block Term Decomposition for the Modelling of Higher-Order Arrays

Characterization of Decomposition of Matrix Multiplication Tensors

Tensor Decompositions for Signal Processing Applications: From two-way to multiway component analysis