Splines Flashcards
1
Q
Splines
A
- piecewise polynomial functions used to approx complex relantionships
- create smooth, flexible curves
- each segment is defined by a polynomial function, with curve combining all segments together
2
Q
Basis Functions
A
- individual polynomial segments
- building blocks to construct the spline
3
Q
Knots
A
- points where the polynomial functions or basis functions join
- choice of number of knots affetcs the flexibility and smoothness of the spline
- more knots - more flexible
- less knots - more smooth
4
Q
Penalised Regression Spline
A
- spline that incorporates a penalty term
- prevents overfitting, ensuring spline captures the underlying trend without being overly sensitive to noise
- tuning parameter - controls trade-off between fit and smoothness
- low lambda - more flexible and wiggly
- high lambda - smoother
5
Q
Smoothness Penalty
A
- measures the integrated squared second derivatve of function, reflecting the curvature of the function
6
Q
Thin-Plate Regression Basis Function
A
- basis function
- depends on Euclidean distance between points
- attach points to thin sheet
7
Q
Why Splines
A
- provide flexible, smooth and interpretable way to capture complex relantionships
- useful for non-linear relantionships to capture the underlying patterns
- clustering - reduce dimension
- spatial - modelling smooth spatial surfaces