Title
Machine Learning with Magnets
Abstract
Machine learning continues to be a broadly popular method of data analysis. Modern machine learning algorithms have enabled data-driven solutions to problems in a diverse set of fields, from computer vision to clinical psychology. These algorithms come in many forms, such as neural networks, kernel methods, and clustering algorithms, to name a few. In this seminar, we will develop the mathematical intuition behind a specific type of neural network, called the Boltzmann Machine. The basic structure of a Boltzmann Machine can be understood in the context of the Ising model, a model of ferromagnetism in statistical mechanics. In developing this understanding, we will see how concepts derived from statistical mechanics are widespread in the mathematics of machine learning.
Powerpoint
Kyle Lennon - Stat Mech Talk.pptx