Instance-based vs Model-based Learning

Instance-based learning

• It makes predictions based on how similar is a new instance to the ones next to it.

• It requires a measure of similarity.

• Examples:

  • k-Nearest Neighbors

  • Kernel Machines

  • Radial Basis Functions Networks (RBF Networks)

Model-based learning

• It "learns" some parameters by a specified performance measure, expressed as:

  • a fitness function: how good the model is, or

  • a cost function: how bad the model is.

• Examples:

  • Linear Regression

  • Logistic Regression

  • Decision trees

  • ...