Example 1: Handwritten Character Recognition
Types of machine learning:
- Supervised learning(Predictive learning) 监督学习
1. Classification
Binary classification
Multiclass classification
Multi-label classification(multiple output model)
2. Applications
Document classification: Internet search, email spam filtering
Image classification: handwritten character recognition, face recognition
Object detection: face detection
- Unsupervised learning(Descriptive learning) 无监督学习
In unsupervised learning, only input data are given without any desired outputs.
Applications:
Discovering clusters
Discovering latent factors and dimensionality reduction
Discovering graph structure
Matrix completion
- Reinforcement learning 强化学习
1. Learning how to act when given occasional reward or punishment signals
2. No supervised answer but delayed reward
3. Credit assignment problem
Why machine learning?
Computers are said to be smart. Why do they need to learn anyway?
- Do not know how to hardwire by hands
Human expertise does not exist (e.g., navigating on Mars)
Humans are unable to explain their expertise (e.g., speech recognit
- Solution changes in time (e.g., routing on a computer network)
- Solution needs to be adapted to particular cases (e.g., user biometrics)