From a94c53e6ce25649827695e7328127958a2903dc8 Mon Sep 17 00:00:00 2001 From: graehl Date: Wed, 28 Jul 2010 16:07:03 +0000 Subject: vest+hope direction non-result git-svn-id: https://ws10smt.googlecode.com/svn/trunk@455 ec762483-ff6d-05da-a07a-a48fb63a330f --- report/vest_mira_hope_direction/README | 9 +++++++++ 1 file changed, 9 insertions(+) create mode 100644 report/vest_mira_hope_direction/README (limited to 'report/vest_mira_hope_direction/README') diff --git a/report/vest_mira_hope_direction/README b/report/vest_mira_hope_direction/README new file mode 100644 index 00000000..c3641f4c --- /dev/null +++ b/report/vest_mira_hope_direction/README @@ -0,0 +1,9 @@ +First column = tune iteration, second column = tune BLEU + +Choosing the average of a few random sentence's model->hope direction, With Vlad's default --bleu_weight=1, nothing improved, and convergenced seemed to be slower, which is the opposite of what we intended. 5 random directions and the orthogonal directions were still used (fewer than normal random directions; the oracle directions replaced some of them). + +dev.bleu.baseline.txt - usual search directions (orth+random) + +dev.bleu.oracle_mira_direction.txt - as above, but fewer random directions, 10 model->hope directions + +note that the search directions are recomputed after each MERT iteration. -- cgit v1.2.3