台大机器学习基石 Lecture 1 - The Learning Problem

 

What is Machine Learning

Defenition:

Improving some performance measure with experience computed from data. 

An alternative route to build complicated systems.

Key Essence:

• exists some ‘underlying pattern’ to be learned
• no programmable (easy) definition
• somehow there is data about the pattern

Key Essence help decide whether to use ML.

Components of ML

Basic Notations:

• input: x ∈ X
• output: y ∈ Y
• unknown pattern to be learned ⇔ target function:    f : X → Y
• data ⇔ training examples: D = {(, ),(, ),··· ,(, )}
• hypothesis ⇔ skill with hopefully good performance:    g: X → Y

Learning Model = A and H

use data to compute hypothesis g that approximates target f