D14-1026 |
correlation with human judgment by
|
hill climbing
|
with 100 random restarts using
|
E09-1059 |
non-trivial we use an approximate
|
hill climbing
|
technique . First we randomly
|
D14-1217 |
results we present . We use a simple
|
hill climbing
|
strategy to find a reasonable
|
D11-1103 |
lattices in the second pass . A
|
hill climbing
|
method ( iterative decoding )
|
E06-2012 |
Alembic name tagger by manual
|
hill climbing
|
. Because this tagger was originally
|
H93-1021 |
likelihood guarantees that any
|
hill climbing
|
method will converge to the global
|
D12-1108 |
test sets demonstrates that the
|
hill climbing
|
decoder manages to fix some of
|
H93-1020 |
Since the training is an iterative
|
hill climbing
|
tech - nique , initialization
|
D12-1108 |
State Initialisation Before the
|
hill climbing
|
decoding algorithm can be run
|
D12-1108 |
Our decoder 's first - choice
|
hill climbing
|
strategy never enumerates the
|
H92-1036 |
estimate of ~ o , we obtain a
|
hill climbing
|
procedure by alternate maximization
|
D12-1108 |
row were obtained by running the
|
hill climbing
|
decoder with DP initialisation
|
D09-1161 |
points ( the Markov link ) for
|
hill climbing
|
. However , it accepts some bad
|
D08-1095 |
based on an error backpropagation
|
hill climbing
|
algorithm ( Diligenti et al.
|
C02-1020 |
correspondence level of the word pairs by
|
hill climbing
|
. These methods could archive
|
E12-1073 |
which can be optimized using e.g.
|
hill climbing
|
. TERp being a tunable metric
|
D11-1103 |
are formed , we follow a similar
|
hill climbing
|
procedure as proposed in our
|
D09-1161 |
temperature ( line 7 - 10 ) . The
|
hill climbing
|
nature gives this algorithm the
|
D12-1108 |
language model . 2 SMT Decoding by
|
Hill Climbing
|
In this section , we formally
|
D12-1108 |
Search Algorithm Parameters The
|
hill climbing
|
algorithm we use has two parameters
|