Information penalty for learning (ECE592 Module 7B)

This module analyzes the penalty for learning an unknown parameter. We do so by designing a (parametric) model class C, and argue for length-N data that C is comprised of (roughly) sqrt(N) quantization levels of the parameter.