监督学习_决策树

Supervised Learning = Approximation--signal exist

Unsupervised Learning = Description-- try to find a signal

density estimation about pixel-- so it is a question about statistics

reinforcement learning is described as learning from delayed reward

generalization -- inductive bias

induction -- from examples to more general rules 

deduction -- opposite

the difference between classification and regression: if the output is discrete then it is the classification

XOR is complicated because the different lines have 2^n then the outcomes have 2^2^n. This is tricky so to understand this, you could try to understand when n=1 and n=2. Then you should be clear.

When the attributes are continuing then it is possible and reasonable to repeat the attribute in the same branch since we may want to know more clearly about the attributes.