Learning Representations from Imperfect Time Series Data via Tensor Rank\n Regularization

2019/07/01 by Paul Pu Liang, Liang, Paul Pu, Zhun Liu +9 · 4 citations
Computer Science · Mathematics · #Computation and Language (cs.CL) #Computational Physics and Python Applications #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Tensor decomposition and applications #Topic Modeling

paper · pdf · doi:10.48550/arxiv.1907.01011

openalex publication_date 2019/07/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

There has been an increased interest in multimodal language processing\nincluding multimodal dialog, question answering, sentiment analysis, and speech\nrecognition. However, naturally occurring multimodal data is often imperfect as\na result of imperfect modalities, missing entries or noise corruption. To\naddress these concerns, we present a regularization method based on tensor rank\nminimization. Our method is based on the observation that high-dimensional\nmultimodal time series data often exhibit correlations across time and\nmodalities which leads to low-rank tensor representations. However, the\npresence of noise or incomplete values breaks these correlations and results in\ntensor representations of higher rank. We design a model to learn such tensor\nrepresentations and effectively regularize their rank. Experiments on\nmultimodal language data show that our model achieves good results across\nvarious levels of imperfection.\n

Cited by

• Partial Information Decomposition via Normalizing Flows in Latent Gaussian Distributions
• MAESTRO : Adaptive Sparse Attention and Robust Learning for Multimodal Dynamic Time Series
• Uncertainty-Aware Collaborative System of Large and Small Models for Multimodal Sentiment Analysis
• Graph-based Interaction Augmentation Network for Robust Multimodal Sentiment Analysis

Related

• Solving Tensor Low Cycle Rank Approximation
• Beyond Low Rank: A Data-Adaptive Tensor Completion Method
• Dictionary Learning with Low-rank Coding Coefficients for Tensor Completion
• Fourier Low-rank and Sparse Tensor for Efficient Tensor Completion
• Theoretical and Experimental Analyses of Tensor-Based Regression and Classification