Closed GoogleCodeExporter closed 8 years ago
[deleted comment]
Look like issue 16.
Original comment by vojta.ry...@gmail.com
on 9 Feb 2012 at 11:47
This issue should now be gone:
valgrind --track-origins=yes bin/rf-ace-build-predictor -I
issue19/Datasets/adult/adult.test.arff -i class -O gbt_issue19.sf
...
==24163==
==24163== HEAP SUMMARY:
==24163== in use at exit: 0 bytes in 0 blocks
==24163== total heap usage: 2,370,461 allocs, 2,370,461 frees, 8,190,464,656
bytes allocated
==24163==
==24163== All heap blocks were freed -- no leaks are possible
==24163==
==24163== For counts of detected and suppressed errors, rerun with: -v
==24163== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 4 from 4)
Original comment by timo.erk...@gmail.com
on 10 Feb 2012 at 10:25
Confirm. (But for given data set rf-ace ran more than 10 minutes and I killed
it. It is weird on such small data...)
Original comment by vojta.ry...@gmail.com
on 11 Feb 2012 at 10:02
True, I noticed as well that the latest changes are oddly making the program
run slower. I will investigate the reason for this behavior.
Original comment by timo.erk...@gmail.com
on 11 Feb 2012 at 11:47
With the latest revision (r463) here are the time consumptions of different
operations:
bin/rf-ace-filter -I issue19/Datasets/adult/adult.test.arff -i class -O
issue19/associations.tsv
...
105.26 seconds elapsed.
bin/rf-ace-build-predictor -I issue19/Datasets/adult/adult.test.arff -i class
-O issue19/gbt.sf
...
7.59 seconds elapsed.
bin/rf-ace-predict -F issue19/gbt.sf -I issue19/Datasets/adult/adult.test.arff
-O issue19/predictions.tsv
...
1.16 seconds elapsed.
NOTE: prediction may give inaccurate results due to partially working forest
(.sf) loader.
Original comment by timo.erk...@gmail.com
on 11 Feb 2012 at 2:40
Original comment by timo.erk...@gmail.com
on 15 Feb 2012 at 3:35
Original issue reported on code.google.com by
vojta.ry...@gmail.com
on 9 Feb 2012 at 11:40Attachments: