aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorGertjan van den Burg <burg@ese.eur.nl>2014-06-26 17:53:51 +0200
committerGertjan van den Burg <burg@ese.eur.nl>2014-06-26 17:53:51 +0200
commit244b5d30b3f794a030be5d943fa9f672e50c38ad (patch)
tree52e645cb278df0a35d140d0e530012037df32a23
parentremove comments from prediction function (diff)
downloadgensvm-244b5d30b3f794a030be5d943fa9f672e50c38ad.tar.gz
gensvm-244b5d30b3f794a030be5d943fa9f672e50c38ad.zip
remove fixed random seed from grid search
-rw-r--r--src/msvmmaj_train_dataset.c4
-rw-r--r--src/trainMSVMMajdataset.c3
-rw-r--r--training/satimage.training1
3 files changed, 1 insertions, 7 deletions
diff --git a/src/msvmmaj_train_dataset.c b/src/msvmmaj_train_dataset.c
index 29b410b..7db1379 100644
--- a/src/msvmmaj_train_dataset.c
+++ b/src/msvmmaj_train_dataset.c
@@ -425,10 +425,6 @@ void consistency_repeats(struct Queue *q, long repeats, TrainType traintype)
* seed_model parameter. If seed_model is NULL, random starting values are
* used.
*
- * @todo
- * There must be some inefficiencies here because the fold model is allocated
- * at every fold. This would be detrimental with large datasets.
- *
* @param[in] model MajModel with the configuration to train
* @param[in] seed_model MajModel with a seed for MajModel::V
* @param[in] data MajData with the dataset
diff --git a/src/trainMSVMMajdataset.c b/src/trainMSVMMajdataset.c
index 35f4611..a6f3e87 100644
--- a/src/trainMSVMMajdataset.c
+++ b/src/trainMSVMMajdataset.c
@@ -98,8 +98,7 @@ int main(int argc, char **argv)
struct Queue *q = Malloc(struct Queue, 1);
make_queue(training, q, train_data, test_data);
- // srand(time(NULL));
- srand(123456);
+ srand(time(NULL));
note("Starting training\n");
if (training->traintype == TT)
diff --git a/training/satimage.training b/training/satimage.training
index bf12e62..d4d08d1 100644
--- a/training/satimage.training
+++ b/training/satimage.training
@@ -1,5 +1,4 @@
train: ./data/satimage.train
-test: ./data/satimage.test
p: 1.0 1.25 1.5 1.75 2.0
kappa: -0.9 0.0 0.5 1.0 5.0
lambda: 64 32 16 8 4 2 1 0.5 0.25 0.125 0.0625 0.03125 0.015625 0.0078125 0.00390625 0.001953125 0.0009765625 0.00048828125 0.000244140625