Lasso (statistics)

Regularization method that tends to cause all but one parameter to be set to 0 during the process of loss minimization. Useful for reducing dimensionality of a model
See Desmos demo for why the above happens

Related

• Ridge (statistics)
• Elastic Nets

${\lambda }_1\sum _{j=1}^d|{\beta }_j|=\lambda \Vert \beta {\Vert }_1$

• This lasso term effectively minimizes the coefficients for any feature $\beta$
• $J$ is the number of parameters
• Effectively uses the L1 norm on each parameter
• This term is added to the loss function