COMPSCI 70 Chapter Notes - Chapter 26: Conditional Expectation, Orthogonality, Random Variable
Document Summary
Goal of llse and linear regression (estimation problem) Llse and linear regression are the best linear fit (linear model) The non-bayesian perspectives treats all data points equally likely. Non-bayesian actually assumes that all sample points are all uniform. Non-bayesian and bayesian model approaches can be studied in one shot (together) Non-bayesian is a special case of the bayesian model. Bayesian is a generalized case of the non-bayesian model. Llse and linear regressions are linear models/version of lse. Because non-bayesian does not assume anything about , it uses the assumption that all data points are equally likely. Non-baysian: having no prior knowledge about the r. v. s and just using just data/sample points, make a linear regression is. Bayesian: having prior knowledge about (we know about the joint distribution of ), make a linear least squares estimate. Goal: find a that minimizes the expected mean squared error, is our best guess given.