ARD is an R package for implementing Automatic Relevance Determination. Automatic Relevance Determination (ARD) is also known as sparse bayesian learning (SBL) and relevance vector machine (RVM). In the context of neural network, the method is called ARD. Applying ARD to neural networks givens an efficient means of variable selection in nonlinear models. ([MacKay, 1992] and [Neal, 1995]). In the context of linear models, this method is called SBL [Tipping, 2001]. Combining ARD and SBL with the use of kernel function in a linear model gives RVM.