跳转至

Lecture7: Learning 3

Decision Tree

ID3 Algorithm

  • If Attributes is empty, return the single-node tree, with the label the most common value of Target_attribute in Examples.

The maximum height of the decision tree is the size of Attributes set.

C4.5

Random Forest

  • Different samples
  • Different features

Bootstrap Sample

  • Drawn with replacement

Construction

  • Assume we have \(n\) examples, \(d\) features

  • Bootstrap samples from training data

  • Construct each decision tree with randomly sampled \(K \approx \sqrt{d}\) features for each spilt

Deep Learning

Deep learning is algorithm that model high-level abstractions in data using architectures consisting of multiple nonlinear transformations.