Spline LASSO in high-dimensional linear regression-荆炳义 (香港科技大学)

A Spline-LASSO approach is proposed for a high-dimensional linear 

regression problem, where the covariates are ordered in a meaningful 

way. It is designed to tackle the case where the shape of the parameter 

values changes smoothly, whereas fused Lasso is better fitted for 

piecewise constant functions. Computationally it can be easily modified 

to use LARS algorithms. Simulations and real example will be given.