Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with
or
.
Download ZIP
Browse files

Add some advice about the regularization coefficient

  • Loading branch information...
commit 57f76c9697d0052a7490d2bc27ecd31d8bb9566d 1 parent 060a0f0
@batterseapower authored
Showing with 6 additions and 0 deletions.
  1. +6 −0 Algorithms/MachineLearning/Framework.hs
View
6 Algorithms/MachineLearning/Framework.hs
@@ -14,6 +14,12 @@ import Numeric.LinearAlgebra
type Target = Double
type Weight = Double
+-- | A positive constant indicating how strongly regularization should be applied. A good
+-- choice for this parameter might be your belief about the variance of the inherent noise
+-- in the samples (1/beta) divided by your belief about the variance of the weights that
+-- should be learnt by the model (1/alpha).
+--
+-- See also equation 3.55 and 3.28 in Bishop.
type RegularizationCoefficient = Double
--
Please sign in to comment.
Something went wrong with that request. Please try again.