Please enable JavaScript.
Coggle requires JavaScript to display documents.
Statistical Learning (Goodness of Fit
MSE= (y(x)-f(x))*2 (Bias Square ( …
Statistical Learning
-
y = f(x) + e
- f is estimated by f*
- f* has following use and types
Prediction
- Mainly function are not interpretable
- Main focus on prediction.
- Ex- SVM, Boosting, Bagging
-
Inference
- Need to find the relationship between response and explanatory variable
- Inference is more important than prediction
3.Regression, Logistic Regression
-
Two Types of f*
Parametric
- Assume the function form
- Estimate the parameter
Easy with small data, prediction not
very exact, can be used with small sample
size. Problem reduces to parameter estmn.
Non Parametric
Need large amount of data to approximate
the form of function. It gives better prediction than parametric.
-
-
-
Types of Study
Three types
-
Semi Supervised Learning
All data don't have
response variable
value. In this we
incorporate both
the data with and without
response variable
-