From 5618cf96060c91d3121d7a420235333d122e47bc Mon Sep 17 00:00:00 2001
From: Alex Rubinsteyn <alex.rubinsteyn@gmail.com>
Date: Tue, 2 Feb 2016 19:00:39 -0500
Subject: [PATCH] results weren't being split by allele
---
.../matrix-completion-hyperparameter-search.py | 13 +++++++------
scripts/create-combined-class1-dataset.py | 4 ++--
2 files changed, 9 insertions(+), 8 deletions(-)
diff --git a/experiments/matrix-completion-hyperparameter-search.py b/experiments/matrix-completion-hyperparameter-search.py
index eaefef90..7fb41891 100755
--- a/experiments/matrix-completion-hyperparameter-search.py
+++ b/experiments/matrix-completion-hyperparameter-search.py
@@ -97,13 +97,13 @@ parser.add_argument(
parser.add_argument(
"--second-hidden-layer-sizes",
- default=[0, 50],
+ default=[0],
type=parse_int_list)
parser.add_argument(
"--dropouts",
- default=[0.0, 0.25],
+ default=[0.0],
type=parse_float_list)
parser.add_argument(
@@ -194,6 +194,7 @@ if __name__ == "__main__":
scores = ScoreSet(
index=[
+ "allele",
"dropout_probability",
"embedding_dim_size",
"hidden_layer_size1",
@@ -250,8 +251,8 @@ if __name__ == "__main__":
embedding_input_dim=21 if args.unknown_amino_acids else 20,
)
predictors[key] = predictor
- initial_weights[key] = predictor.model.get_weights()
- initial_optimizer_states[key] = predictor.model.optimizer.get_state()
+ initial_weights[key] = predictor.model.get_weights().copy()
+ initial_optimizer_states[key] = predictor.model.optimizer.get_state().copy()
# want at least 5 samples in each fold of CV
# to make meaningful estimates of accuracy
@@ -368,7 +369,7 @@ if __name__ == "__main__":
training_sample_weights = 1.0 / np.array(training_counts)
Y_train = np.array([
train_dict[p] for p in training_row_peptides])
- for key, predictor in predictors.items():
+ for key, predictor in sorted(predictors.items()):
print("\n-----")
print(
@@ -400,7 +401,7 @@ if __name__ == "__main__":
y_pred=y_pred,
max_ic50=args.max_ic50)
scores.add_many(
- key,
+ ("%s," % allele) + key,
mae=mae,
tau=tau,
f1_score=f1_score,
diff --git a/scripts/create-combined-class1-dataset.py b/scripts/create-combined-class1-dataset.py
index 486e8fe3..094b6f5b 100755
--- a/scripts/create-combined-class1-dataset.py
+++ b/scripts/create-combined-class1-dataset.py
@@ -137,7 +137,7 @@ if __name__ == "__main__":
for (allele, count) in new_allele_counts.most_common():
print("%s: %d" % (allele, count))
print("Combined DataFrame size: %d (+%d)" % (
- len(combined_df),
- len(combined_df) - len(nielsen_data)))
+ len(combined_df),
+ len(combined_df) - len(nielsen_data)))
print("Writing %s..." % args.output_csv_path)
combined_df.to_csv(args.output_csv_path, index=False)
--
GitLab