[01211] Generalized and non-Gaussian Tensor Decompositions
Session Date & Time :
01211 (1/2) : 1C (Aug.21, 13:20-15:00)
01211 (2/2) : 1D (Aug.21, 15:30-17:10)
Type : Proposal of Minisymposium
Abstract : Tensor decompositions are a foundational unsupervised machine learning method for data science, with applications across all of science and engineering. Traditionally, tensor decompositions seek low-rank tensors that best fit the data with respect to the least squares loss. However, other choices of loss function can be more appropriate for non-Gaussian data such as count data, binary data, and data with outliers. This minisymposium presents state-of-the-art advances in developing efficient algorithms and rigorous theory for tensor decompositions with respect to general losses.
