Data convergencechange development dopar_multicore expgradcpp gh-pages hess_subsample manualinit master multicore pr/27 weights
Nothing to show
Commits on Aug 15, 2017
Aug 15, 2017
- when ngroups > 1, randomly assign indices to groups so that the batches are representative even if the documents are in a non-random order - in memoization, randomly order the batches in each pass (as in Huges and Sudderth, http://cs.brown.edu/~sudderth/papers/nips13dpMemoized.pdf, p. 4.
Commits on Dec 9, 2016