def setUp(self):
folder = dirname(__file__)
files = [
folder + "/data/dna_pos.fasta", folder + "/data/dna_neg.fasta"
]
self.data = Data(files, "ACGT")
self.params = {'conv_num': [1], 'kernel_num': [2, 4], 'epochs': [1, 2]}
self.searcher = Grid_Search(self.params)
class Test_Grid_Search(unittest.TestCase):
def setUp(self):
folder = dirname(__file__)
files = [
folder + "/data/dna_pos.fasta", folder + "/data/dna_neg.fasta"
]
self.data = Data(files, "ACGT")
self.params = {'conv_num': [1], 'kernel_num': [2, 4], 'epochs': [1, 2]}
self.searcher = Grid_Search(self.params)
def test_grid_search_init(self):
self.assertTrue(self.searcher.params == self.params)
self.assertTrue(len(self.searcher.candidates) == 4)
self.assertTrue({
'conv_num': 1,
'kernel_num': 2,
'epochs': 1
} in self.searcher.candidates)
self.assertTrue({
'conv_num': 1,
'kernel_num': 2,
'epochs': 2
} in self.searcher.candidates)
self.assertTrue({
'conv_num': 1,
'kernel_num': 4,
'epochs': 1
} in self.searcher.candidates)
self.assertTrue({
'conv_num': 1,
'kernel_num': 4,
'epochs': 2
} in self.searcher.candidates)
def test_grid_search_train(self):
# filter tensorflow deprecation warnings to not clutter the unittest output
with warnings.catch_warnings():
warnings.simplefilter("ignore")
model, table = self.searcher.train(self.data, verbose=False)
self.assertTrue(isinstance(model, Model))
self.assertTrue(isinstance(table, str))
table = table.split('\n')
self.assertTrue(len(table) == 9)
self.assertTrue(
table[0] in
["# conv_num: [1]", "# kernel_num: [2, 4]", "# epochs: [1, 2]"])
self.assertTrue(
table[1] in
["# conv_num: [1]", "# kernel_num: [2, 4]", "# epochs: [1, 2]"])
self.assertTrue(
table[2] in
["# conv_num: [1]", "# kernel_num: [2, 4]", "# epochs: [1, 2]"])
for word in table[3].split():
self.assertTrue(
word in ["conv_num", "kernel_num", "epochs", "roc-auc"])
for line in table[4:8]:
self.assertTrue(len(line.split()) == 4)
self.assertTrue(table[8] == '')
data.load_additional_data(indel_len_feat, is_categorical=False)
print(data.get_summary())
data.train_val_test_split(portion_train=0.6, portion_val=0.2, seed=3)
print(data.get_summary())
###Model Training
params = {
"conv_num": [2, 3],
"kernel_num": [100],
"kernel_len": [8],
"dropout_input": [0.1, 0.4]
}
searcher = Grid_Search(params)
start = time()
model, summary = searcher.train(data, pr_auc=True, verbose=False)
stop = time()
print("time in minutes: {}".format((stop - start) / 60))
print(summary)
##Perfomance evaluation
predictions = model.predict(data, "test")
predictions
labels = data.get_labels("test")
labels
utils.plot_roc(labels, predictions, output_folder + "roc.png")