Dale Smith, Data Scientist, Nexidia
summary
Explains how tensors generalize matrices and that practical decompositions are sum of rank one tensors (CP) or multilinear PCA, shows libraries to run them including Tensor Toolbox (MATLAB), t++ with Python bindings, R tensor and several Python packages, and highlights applications in recommender systems, time‑evolving networks and topic modeling. Makes the concrete claim that spectral tensor methods yield a guaranteed global minimum for many latent variable models, avoid NP‑hard matrix formulations and can be made embarrassingly parallel.
AI-written summary · about
chapters (6)
- 0:00Start
- 0:10Singular Value Decomposition
- 1:16What Is a Tensor
- 4:54Topic Modeling
- 7:45Guaranteed Learning
- 9:06Latent Variable Models
related talks
01h 03mRecent Progress in High-Dimensional Learning
58mTensor Decomposition, Sparse Representations and Applications
56mTensor Decompositions for Estimating Latent Variable Models
52mTensor Methods for Learning Latent Variable Models: Theory and Practice
07mTamara Kolda, Distinguished Member of Technical Staff, Sandia National Laboratories @ MLconf SF Part
02mTamara Kolda, Distinguished Member of Technical Staff, Sandia National Laboratories @ MLconf SF Part
07mWTT: What the Tensor?
29mLarge-Scale Matrix Operations