Recitation 4 M ay 23
description
Transcript of Recitation 4 M ay 23
![Page 1: Recitation 4 M ay 23](https://reader036.fdocuments.in/reader036/viewer/2022070415/56814e80550346895dbc1c69/html5/thumbnails/1.jpg)
RECITATION 4MAY 23
DPMM
Splines with multiple predictors
Classification and regression trees
![Page 2: Recitation 4 M ay 23](https://reader036.fdocuments.in/reader036/viewer/2022070415/56814e80550346895dbc1c69/html5/thumbnails/2.jpg)
Dirichlet Process Mixture Model
• Library “DPpackage”
• R Demo 1
![Page 3: Recitation 4 M ay 23](https://reader036.fdocuments.in/reader036/viewer/2022070415/56814e80550346895dbc1c69/html5/thumbnails/3.jpg)
Spline method with multiple predictors
• Generalized Additive Model
• Natural Thin Plate Splines• The minimizer of (RSS+“bending energy”) among all interpolators
with knots at the observations.
• Form:
![Page 4: Recitation 4 M ay 23](https://reader036.fdocuments.in/reader036/viewer/2022070415/56814e80550346895dbc1c69/html5/thumbnails/4.jpg)
Spline method with multiple predictors
• Thin Plate Regression Splines• Optimal approximation of thin plate splines using low rank basis• No need to choose knots
• Tensor Product Splines• Basis: product of basis (truncated spline) of each dimension
• R Demo 2
![Page 5: Recitation 4 M ay 23](https://reader036.fdocuments.in/reader036/viewer/2022070415/56814e80550346895dbc1c69/html5/thumbnails/5.jpg)
Classification and regression trees
• Classification tree• The response is binary or categorical outcome.
• Regression tree• The response is a continuous variable. The predicted value will be
the same for all data points in a leaf node.
• “Grow” the tree and then “prune” it by minimizing cross validation error
• R Demo 3
![Page 6: Recitation 4 M ay 23](https://reader036.fdocuments.in/reader036/viewer/2022070415/56814e80550346895dbc1c69/html5/thumbnails/6.jpg)
Course Evaluation• Thanks!