Message1455

Author malte
Recipients malte, moritz
Date 2011-08-04.19:19:12
Content
Makes sense. It also makes sense that this would show up primarily in
ParcPrinter, since this has solution costs in the range of millions while other
domains are almost always below 1000.

I think that the natural generalization of DFP to the case where there are M
distinct h values and N < M abstract states is to allocate N - 1 abstract states
to the lowest N - 1 h values and throw all the larger ones into a single
abstract state.

If it helps avoid code duplication (not sure if it does), the existing
bucket-based shrinking code could be used for that if we have one that considers
only the h value, not the f value. (I thought he had one like this, but now I
can't find it any more.)
History
Date User Action Args
2011-08-04 19:19:12maltesetmessageid: <1312478352.29.0.518054019975.issue250@gmail.com>
2011-08-04 19:19:12maltesetrecipients: + malte, moritz
2011-08-04 19:19:12maltelinkissue250 messages
2011-08-04 19:19:12maltecreate