creating quadratic features for pairs: ui 
using l2 regularization = 2e-06
Num weight bits = 16
learning rate = 0.05
initial_t = 1
power_t = 0
decay_learning_rate = 0.97
rank = 10
final_regressor = movielens.reg
creating cache_file = movielens.cache
Reading datafile = train-sets/ml100k_small_train
num sources = 1
average    since         example     example  current  current  current
loss       last          counter      weight    label  predict features
3.457930   3.457930            1         1.0   2.0000   0.1404       23
5.743041   8.028152            2         2.0   3.0000   0.1666       23
9.391515   13.039988           4         4.0   4.0000   0.5033       23
9.041728   8.691941            8         8.0   2.0000   0.8482       23
6.344679   3.647631           16        16.0   3.0000   1.3200       23
4.815768   3.286856           32        32.0   2.0000   2.1091       23
3.394970   1.974173           64        64.0   5.0000   2.8885       23
2.366933   1.338895          128       128.0   4.0000   3.8108       23
1.826340   1.285746          256       256.0   3.0000   3.1742       23

finished run
number of examples per pass = 237
passes used = 2
weighted example sum = 474
weighted label sum = 1666
average loss = 1.51233
best constant = 3.52008
total feature number = 10902