Explore formal models of learnability in this 48-minute machine learning lecture focusing on batch learning and computational learning theory. Delve into the mistake bound model before examining what formal models of learnability provide and understanding the goals of batch learning. Learn about learning conjunctions, random source training examples, nature-provided labels, and different approaches to evaluating learning algorithms. Understand key concepts including distribution over instance space, PAC Learning intuition, hypothesis error, empirical error, and the distinctions between online and batch learning methodologies.
Machine Learning: Computational Learning Theory and Batch Learning - Lecture 12
Overview
Syllabus
Intro
Checkpoint: The bigger picture
Learning Conjunctions Some random source (nature) provides training examples Teacher (Nature) provides the labels (f(x))
Two Directions for How good is our learning algorithm?
The mistake bound approach
The setup
Distribution over the instance space
PAC Learning - Intuition
Error of a hypothesis
Empirical error
The goal of batch learning
Online learning vs. Batch learning
Taught by
UofU Data Science
