3. Types of learning
P10 3.1
There are only 2 types of output: binary classification (true or false questions)
the machine outputs real numbers for questions: regression analysis/regression analysis with range
Different y:
structured learning: y = structures natural language processing, protein display
P11 3.2
Supervised learning: Give {xi,yi}
1 Supervised learning
2 Unsupervised learning (unsupervised): only give xi, and then the machine itself clusters (clustering clustering), so it may be wrong
Use: See Figure
3 Semi-supervised: Mark Some points
4 reinforcement learning: similar to dog training,
tell the output of the machine whether it is good or not to learn, and train batch by batch
Introduction to unsupervised learning:
P12 3.3
Summary:
1 batch: feed a batch of data to the machine
2 online:
garbage Mailbox identification: one by one email identification: online learning
is continuously updated w
3 active learning: the machine asks questions, used when obtaining labels is very expensive
P13 3.4
Data input:
1 specific data, machine learning difficulty is low
2 not specific , learning difficulty becomes higher
3 Abstract data, learning difficulty is the most difficult
Finally, a small summary of Chapter 3:
