Seminar, Hosted by Institute for Computational and Experimental Research in Mathematics (ICERM), Brown University
Matrix-Mimetic Tensor Algebra: Optimal Decompositions and Equivariant Learning
Abstract
Lior Horesh presents tensor-tensor algebra designed to retain key properties of matrix algebra while representing multidimensional correlations. An Eckart–Young-like tensor representation theorem underpins computationally feasible, provably optimal decompositions. Matrix-mimetic operations allow existing computational workflows to be adapted to tensors. Examples include tensorized neural-network structures and tensor graph convolutional networks for time-evolving graphs. The discussion concludes with tensor group symmetries and extensions to equivariant learning.
Topics
- tensor-tensor algebra, broader topic: algebra
- multidimensional correlations, broader topic: correlations
- Eckart–Young-like tensor representation theorem, search World Wide
- Matrix-mimetic operations, search World Wide
- tensor graph convolutional networks, broader topic: convolutional networks
- time-evolving graphs, broader topic: graphs
- tensor group symmetries, broader topic: symmetries
- equivariant learning, broader topic: learning