Had a minor refactor of feature normalization related code. Changes are:
Fixed a bug in calculating variance when feature normalization is enabled. Previous variance calculation function createModel was wrong because means and variance cannot share the same transformation. The correct way of calculating variance when feature normalization is enabled is: first transform the means back to the original space, and then call computeVariances function.
Added normalizationContext in PriorDistribution.scala to enable mean and variance normalization in incremental learning.
Had a minor refactor of feature normalization related code. Changes are:
Fixed a bug in calculating variance when feature normalization is enabled. Previous variance calculation function
createModel
was wrong because means and variance cannot share the same transformation. The correct way of calculating variance when feature normalization is enabled is: first transform the means back to the original space, and then callcomputeVariances
function.Added
normalizationContext
inPriorDistribution.scala
to enable mean and variance normalization in incremental learning.