KrisLibrary
1.0.0
|
A linear model of n-dimensional data. More...
#include <LinearModel.h>
Public Types | |
enum | SolveMethod { Cholesky, SVD, QR } |
Public Member Functions | |
Real | Evaluate (const Vector &x) const |
Real | Variance (const Vector &x) const |
void | CoeffVariance (Vector &coeffVariance) const |
bool | LeastSquares (const Matrix &x, const Vector &outcome) |
bool | LeastSquares (const std::vector< Vector > &x, const std::vector< Real > &outcome) |
bool | LeastSquares (const std::vector< Vector > &x, int outcomeIndex) |
Public Attributes | |
SolveMethod | solveMethod |
bool | constantOffset |
Vector | coeffs |
Matrix | covariance |
A linear model of n-dimensional data.
If constantOffset = true, there are n+1 coefficients, and the n+1'th element of coeffs is a constant offset. That is, the model is x^T coeffs(0:n-1) + coeffs(n). Otherwise, the model is x^T coeffs.
LeastSquares assumes normally distributed data with independent errors.