inputs corresponding to observations

vector of penalty parameters lambda

error adjustment to majorization function

maximum number of iterations allowed

max_iter

Uses majorization-minimization technique to approximate trend filtering fit.

Trend filtering uses the generalized
lasso framework to fit an adaptive polynomial of degree k to
estimate the function f_0 at each input x_i in the model: y_i =
f_0(x_i) + epsilon_i, for i = 1, ..., n, and epsilon_i
is sub-Gaussian with E(epsilon_i) = 0.  Bayesian trend filtering adapts
the genlasso framework to a fully Bayesian hierarchical model, estimating
the penalty parameter lambda within a tractable Gibbs sampler.

tf: approximate trend filtering via MM algorithm

Description

Usage

Arguments