Lecture 20: Machine Learning Framework, Unsupervised Machine Learning and Its Quantum Extension скачать в хорошем качестве

Lecture 20: Machine Learning Framework, Unsupervised Machine Learning and Its Quantum Extension

Support vector structure for identifying the margin in binary classification is explained. Cost functions for primal and dual formulations, and regression problems, are described. Neural network models with neurons as binary variables are defined, whose couplings are tuned in reinforcement learning. Unsupervised machine learning with Boltzmann distribution is explained. It minimises the relative entropy of distributions in both classical and quantum versions. Connection to the Fisher information metric and the Cramer-Rao bound is pointed out. Quantum advantage is possible when the quantum state and the observable of interest do not commute. For a single qubit, maximum quantum advantage is obtained when the density matrix is transverse to the observable. The assignments for the course can be found here: https://chep.iisc.ac.in/Personnel/adp...