Smoothed analysis of tensor decompositions

@inproceedings{Bhaskara2014SmoothedAO,
  title={Smoothed analysis of tensor decompositions},
  author={Aditya Bhaskara and M. Charikar and A. Moitra and Aravindan Vijayaraghavan},
  booktitle={STOC '14},
  year={2014}
}
  • Aditya Bhaskara, M. Charikar, +1 author Aravindan Vijayaraghavan
  • Published in STOC '14 2014
  • Computer Science, Mathematics
  • Low rank decomposition of tensors is a powerful tool for learning generative models. The uniqueness results that hold for tensors give them a significant advantage over matrices. However, tensors pose serious algorithmic challenges; in particular, much of the matrix algebra toolkit fails to generalize to tensors. Efficient decomposition in the overcomplete case (where rank exceeds dimension) is particularly challenging. We introduce a smoothed analysis model for studying these questions and… CONTINUE READING
    108 Citations

    Topics from this paper.

    Tensor decompositions for learning latent variable models
    • 788
    • Highly Influenced
    • PDF
    Optimal Low-Rank Tensor Recovery from Separable Measurements: Four Contractions Suffice
    • 12
    • Highly Influenced
    • PDF
    Tensor Networks for Latent Variable Analysis: Higher Order Canonical Polyadic Decomposition
    • 7
    • PDF
    Tensor Completion Made Practical
    • 2
    • PDF
    A Robust Spectral Algorithm for Overcomplete Tensor Decomposition
    • 5
    • PDF
    Effective Criteria for Specific Identifiability of Tensors and Forms
    • 35
    • PDF
    Fast and robust tensor decomposition with applications to dictionary learning
    • 21
    • PDF
    Sparse and Low-Rank Tensor Decomposition
    • 18
    • Highly Influenced
    • PDF
    Learning Overcomplete Latent Variable Models through Tensor Methods
    • 50
    • PDF

    References

    SHOWING 1-9 OF 9 REFERENCES
    Subtracting a best rank-1 approximation may increase tensor rank
    • A. Stegeman, P. Comon
    • Mathematics, Computer Science
    • 2009 17th European Signal Processing Conference
    • 2009
    • 51
    • Highly Influential
    • PDF
    Most Tensor Problems Are NP-Hard
    • 674
    • Highly Influential
    • PDF
    A Tensor Spectral Approach to Learning Mixed Membership Community Models
    • 106
    • Highly Influential
    • PDF
    A Spectral Algorithm for Latent Dirichlet Allocation
    • 252
    • Highly Influential
    • PDF
    Independent component analysis, A new concept?
    • P. Comon
    • Mathematics, Computer Science
    • Signal Process.
    • 1994
    • 7,860
    • Highly Influential
    • PDF
    Three-way arrays: Rank and uniqueness of tri-linear decompositions
    • 159
    • Highly Influential
    Foundations of the PARFAC procedure: model and conditions for an ‘explanatory’ multi-mode factor analysis
    • UCLA Working Papers in Phonetics, pages 1–84,
    • 1970
    Fourier PCA
    • 19
    • Highly Influential
    Smoothed Analysis of Tensor Decompositions
    • arxiv:1311.3651,
    • 2013