CMU: Intro to Ml Notes

第一课

Artificial Intelligence (AI): Getting Computers to Behave Intelligently

Do things that people do well (better than computers) A moving target! But usually involves:

• Perceptron
• Control
• Planning
• Human language (recognize, understand, respond to, generate)

Example Tasks:

• Identify objects in an image
• Translate from one huamn language to another
• Recognoize speech
• Assess risk (e.g. in loan application)
• …

1st Attempt: “Knowledge-Based AI” (1960s — 1980s)

= Write programs that simulate how people do it. Problems:

• Will never get better than a person
• Requires deep introspection
• Sometimes requires experts (“expert systems”, “knowledge elicitation”)
• Often we don’t know how we do things (e.g. ride bicycle)
  • Difference between knowing and knowing-how-we-know
• Sometimes we think we know, but we’re wrong

Didn’t work as well as hoped.

Alternative: “Data-Based AI” (a.k.a Machine Learning) (1980s — today)

= Write programs that learn the task from examples.

• You don’t need to know how to do it yourself
• Performance (should) improve with more examples

But:

• Need lots of examples! (millions and billions)
• When it finally works, you may not understand how

Lecture 2

ML: Learn a task from experience (examples)

• Get better at the task with experience

Medical Diagnosis: f: Patient Info -> {Categories}, Classification

Predict Tomorrow’s Temperature f: (Input) -> R, Regular Regression

ML = Learning a function.