todo.txt

Non-parametric density estimation

==================================================================
Stuff to Fix, Change, or Add:
==================================================================

"EvalAvgLogL" function (smart version)

Higher order "multipole-like" expansion?  Set order p somehow?

llGrad -> other gradients (bw, etc)
       -> entropy, kl, mi accessor f'ns.
       -> Hessian of log-likelihood
           => newton-like search methods

Distance-based KL-est stuff

Max Log-Error function  (add to other error measures: KL, ISE)
  KL error : override expectation distribution ( E_p log p1/q1 )

Add KD-tree bound-based reduction method to "reduce"

Profiling

Math / uP optimization flags