ML Lecture 0: Introduction of Machine Learning

1. AI v.s. ML v.s. DL
• Artificial intelligence -> objective
• Machine learning -> methods
• Deep learning -> one method of machine learning
2. Hand-crafted rules
   • many "ifs"
   • hard to consider all possibilities
   • no learning (limited)
   • lots of human efforts (not suitable for small industry)
3. Machine learning
   • write programs for learning
   • looking for a function from data
     • 1. defining a set of functions -> model
     • 2. training data -> evaluate the goodness of functions (supervised learning)
     • 3. picking the best function f star
       
     • 4. using f star
   • 1-3 -> traing 4 -> testing
4. Learning map
   • 
5. Regression (output: scalar/value/real number)
   • e.g. prediction of PM2.5
6. Classification
   • binary classification (e.g. spam filtering)
   • multi-class classification (e.g. document classification)
7. Supervised learning
   • training data: input / output pair of target function (function output -> label)
   • hard to collect a large amount of labelled data -> semi-supervised learning / tramsfer learning / unsuperviesd learning / reinforcement learning
8. Structured learning - beyond classification
   • e.g. speech recognition / machine translation
9. Reinforcement learning
   • no correct answers but only critics (評價)

李宏毅機器學習 - 0 Introduction