# -*- mode: org -*-
#+STARTUP: lognotestate
#+SEQ_TODO: MAYBE(m) TODO(t) NEXT(n) STARTED(s) WAITING(w) | DONE(d) CANCELLED(c)
#+STARTUP: hidestars
* unroll only part of the network
* example for unrolling with missing values
* factored rbm
* higher order rbm
* SRBM
* true batch training (for speed)
* clarify conjugate gradient license
* cg: make sbcl specific float code portable
* bp: the structure is static so there is no point in ->+ (&REST ARGS)
* faster exp? http://citeseer.ist.psu.edu/cache/papers/cs/43/ftp:zSzzSzftp.idsia.chzSzpubzSzniczSzexp.pdf/schraudolph98fast.pdf
Doesn't seem to speed up things.
* cg: failed line searches can still change the weights
document it or fix it
* rbm: sparsity constraint
* rbm: generic support for exponential family distributions
http://www.ics.uci.edu/~michal/GenHarm3.pdf