Tensor Linear Algebra: The Next Dimension
Join the Libraries and the Mathematics Department at a Young Academic Alumni Lecture Series talk by Elizabeth Newman '14.
Monday, December 9, 2019
Tea at 3:30pm in KINSC H208
Talk at 4:00pm in KINSC H109
Student lunch with Dr. Newman from 12:00pm to 1:00pm in DC 106 (Smith Room), meal tickets provided
As data have become more complex and high-dimensional to reflect the real world (e.g., image and video data), there has been great interest in extending linear algebra to handle these multidimensional data (i.e., tensors) efficiently. However, this extension to multilinear algebra is non-trivial because many fundamental matrix concepts break down when adding new dimensions. After introducing the singular value decomposition, we'll outline some popular methods to extend this concept to tensors, highlighting the inconsistencies with traditional linear algebra that can arise. The main focus of the talk will be discussing a matrix-mimetic, tensor algebra based on a class of tensor-tensor products, showing how this can outperform traditional matrix approaches through some numerical examples. This talk should be understandable to anyone who has taken linear algebra.
Dr. Newman graduated from Haverford College in 2014 where she majored in mathematics and statistics and played varsity women's soccer and softball (Go Fords!). She graduated with a Ph.D in mathematics from Tufts University in May 2019. She now is a visiting assistant professor at Emory University working on scientific machine learning.