Tensor Topic Modeling via HOSVD
A topic-model estimator that generalizes probabilistic latent semantic indexing from a document-by-word matrix to a document-by-word-by-context tensor (adding temporal, spatial, or document-specific modes) by fitting a Tucker decomposition through a modified higher-order singular value decomposition. Students learn how the core tensor captures interactions among latent factors, how the method recovers low-rank topic and cluster structure across multiple dimensions, and how its estimation error is bounded by an explicit entry-wise l1 convergence rate.
Tensor Topic Modeling Via HOSVD
By representing documents as mixtures of topics, topic modeling has allowed the successful analysis of datasets across a wide spectrum of applications ranging from ecology to genetics. An important b…