Message6472

Author jendrik
Recipients jendrik, malte
Date 2017-08-31.11:00:58
Content
The results are in and I think they look better than expected:

http://ai.cs.unibas.ch/_tmp_files/seipp/issue731-v4-opt-issue731-base-issue731-v4-compare.html
http://ai.cs.unibas.ch/_tmp_files/seipp/issue731-v4-sat-issue731-base-issue731-v4-compare.html

(The unexplained errors for the satisficing experiment are due to VAL running out of memory. We should 
probably discuss what to do about that in our next Fast Downward meeting.)

As usual we're seeing quite some noise: blind search is among the configurations with the highest 
overall difference for score_search_time (-1.37), but I don't see any place where new hash functions 
are used in this configuration. 

The most critical configuration is probably the one using systematic PDBs. It's total score_total_time 
decreases by 3.41.

What are your thoughts on this?
History
Date User Action Args
2017-08-31 11:00:58jendriksetmessageid: <1504170058.76.0.0104906573067.issue731@unibas.ch>
2017-08-31 11:00:58jendriksetrecipients: + jendrik, malte
2017-08-31 11:00:58jendriklinkissue731 messages
2017-08-31 11:00:58jendrikcreate