Abstract
This paper investigates the solution of the feature selection problem in biochemical signal transduction pathways by examining the sensitivity of the features with respect to the model complexity using basis pursuit regularization (BPR). Feature selection is effectively transformed into a continuous regularization problem with a characteristic 1-norm imposed on the parameter vector to penalize the models complexity. This technique makes possible the design of sparse models for the pathway data and because of the nature of the 1-norm it is possible to analyze the entire solution path (parameter locus) as the regularizer changes from zero to infinity.
Index
Terms
Available to subscribers and IEEE members.
References
Available to subscribers and IEEE members.
Citing Documents
Available to subscribers and IEEE members.
Cart
