diff options
| author | John Langford <jl@nyclamp.(none)> | 2013-01-14 23:09:19 +0400 |
|---|---|---|
| committer | John Langford <jl@nyclamp.(none)> | 2013-01-14 23:09:19 +0400 |
| commit | e811bb8f7ae719784ee01298cc7fbfdcf4df169c (patch) | |
| tree | 6d0c719c80a9b670db4e269d958e78ebd546b90d /test/train-sets/ref/rcv1_small.stderr | |
| parent | cce2673ffc1e63e7e79f8d083503c94a53114221 (diff) | |
update tests for fix to bfgs
Diffstat (limited to 'test/train-sets/ref/rcv1_small.stderr')
| -rw-r--r-- | test/train-sets/ref/rcv1_small.stderr | 35 |
1 files changed, 13 insertions, 22 deletions
diff --git a/test/train-sets/ref/rcv1_small.stderr b/test/train-sets/ref/rcv1_small.stderr index 86518b8a..0677ece6 100644 --- a/test/train-sets/ref/rcv1_small.stderr +++ b/test/train-sets/ref/rcv1_small.stderr @@ -1,40 +1,31 @@ enabling BFGS based optimization **without** curvature calculation +m = 7 +Allocated 72M for weights and mem +## avg. loss der. mag. d. m. cond. wolfe1 wolfe2 mix fraction curvature dir. magnitude step size time Num weight bits = 20 learning rate = 0.5 initial_t = 1 power_t = 0.5 decay_learning_rate = 1 using l2 regularization = 1 -m = 7 -Allocated 72M for weights and mem -## avg. loss der. mag. d. m. cond. wolfe1 wolfe2 mix fraction curvature dir. magnitude step size time creating cache_file = train-sets/rcv1_small.dat.cache Reading from train-sets/rcv1_small.dat num sources = 1 - 1 0.69315 0.01247 109.37472 165.25444 1072660.62500 0.66186 0.295 - 3 24.96380 0.09448 58.49783 -0.335275 -0.667864 (revise x 0.5) 0.33093 0.342 - 4 6.68890 0.07916 21.38882 -0.165651 -0.341257 (revise x 0.5) 0.16546 0.403 - 5 2.02595 0.05334 9.61256 -0.073645 -0.172136 (revise x 0.5) 0.08273 0.463 - 6 0.88609 0.02271 4.21489 -0.021323 -0.077380 (revise x 0.5) 0.04137 0.523 - 7 0.65787 0.00567 2.42293 0.007797 -0.022375 (revise x 0.5) 0.02068 0.583 - 8 0.64090 0.00109 2.53848 0.023094 0.007617 5.59308 1.00000 0.751 - 9 0.59157 0.00084 1.98472 0.939814 0.880475 449.76361 1.00000 0.944 -10 0.36198 0.00015 0.22195 0.590336 0.255459 123.06779 1.00000 1.169 -11 0.34818 0.00506 0.68004 0.218960 -0.427093 27.43001 1.00000 1.426 -12 0.32328 0.00029 0.08338 0.597312 0.217825 3.59904 1.00000 1.717 -13 0.31983 0.00001 0.04912 0.738423 0.480952 8.91247 1.00000 2.037 -14 0.31597 0.00006 0.04996 0.762549 0.528469 24.29697 1.00000 2.391 -15 0.31083 0.00018 0.05458 0.726800 0.455732 103.02180 1.00000 2.746 -16 0.29800 0.00004 0.01069 0.692401 0.384374 50.95761 1.00000 3.099 -17 0.29486 0.00000 0.00018 0.516695 0.031777 0.39479 1.00000 3.455 -18 0.29479 0.00000 0.00009 0.596026 0.192467 0.24629 1.00000 3.810 + 1 0.69315 0.00186 3.76860 34.14409 19774.77734 0.11037 0.213 + 3 0.46246 0.01086 1.93693 0.554592 0.194583 220.86601 1.00000 0.265 + 4 0.33845 0.00042 0.17266 0.520733 0.133178 34.15247 1.00000 0.332 + 5 0.31850 0.00008 0.06344 0.751656 0.517330 67.75451 1.00000 0.408 + 6 0.30246 0.00000 0.01461 0.657158 0.328843 36.18754 1.00000 0.493 + 7 0.29676 0.00000 0.00352 0.645592 0.293725 15.85137 1.00000 0.584 + 8 0.29527 0.00000 0.00163 0.527831 0.052654 5.80239 1.00000 0.685 + 9 0.29506 0.00000 0.00155 0.199875 -0.608862 0.73891 1.00000 0.796 finished run number of examples = 200000 weighted example sum = 2e+05 weighted label sum = -1.272e+04 -average loss = 0.4485 +average loss = 0.2911 best constant = -0.06361 total feature number = 15587880 -19 0.29476 0.00000 0.00003 0.619462 0.238168 0.15641 1.00000 4.188 +10 0.29483 0.00000 0.00012 0.585417 0.171988 0.14531 1.00000 1.029 |