Tamara Kolda |
Mathematical Consultant
MathSci.ai |
Tech Vision Talk: A Mathematician's View of Machine Learning (and Why It Matters)
Abstract: A (trained) machine learning model, such as a deep neural network, operates loosely as follows: it takes features as an input and produces a classification as an output. The training phase of machine learning tunes the parameters of the machine learning method, e.g., the weights and biases of a neural network, by considering a training set of features with known outputs. Mathematicians think about machine learning models as parameterized functions. From this viewpoint, we can trace the evolution of machine learning over the last few decades, considering what has changed from a mathematical point of view and phenomena such as “double descent”. The mathematical point of view offers an alternative to anthropomorphizing “learning” and enables us to better understand both the workings and the boundaries of ML methods. We can argue that “more data” and “bigger models” are not a panacea, and instead develop mathematical methodology for understanding how to move beyond the current limits of machine learning.
|
Biography
|
Tammy is an independent mathematical consultant under the auspices of her company MathSci.ai based in California. From 1999-2021, she was a member of the technical staff at Sandia National Laboratories, California. She specializes in mathematical algorithms and computation methods for tensor decompositions, tensor eigenvalues, graph algorithms, randomized algorithms, machine learning, network science, numerical optimization, and distributed and parallel computing. She is the founding editor-in-chief for the SIAM Journal on Mathematics of Data Science. She is a member of the National Academy of Engineering (NAE), Fellow of the Society for Industrial and Applied Mathematics (SIAM), and Fellow of the Association for Computing Machinery (ACM).
|