def get_reults(self, bellwethers=None, n_reps=12):
data_path = os.path.realpath("./data")
projects = get_all_projects(data_path)
for project in projects:
print(project.name.upper())
files = project.files()
results_0 = []
for transfer in [Pooyan, Baseline, Waterloo]:
results_1 = [transfer.__doc__.upper()]
for source_name, source_conf in files.iteritems():
for target_name, target_conf in files.iteritems():
if not source_name == target_name:
if transfer.__doc__ == "baseline":
if source_name in bellwethers[project.name]:
results_1.extend([
transfer.learner(
source_conf, target_conf)
for _ in xrange(n_reps)
])
else:
results_1.extend([
transfer.learner(source_conf, target_conf)
for _ in xrange(n_reps)
])
results_0.append(results_1)
rdivDemo(results_0)
print("")
def main(n_reps=30):
data_path = os.path.realpath("./data")
projects = get_all_projects(data_path)
results = dict()
for project in projects:
files = project.files()
results_0 = dict()
for source_name, source_conf in files.iteritems():
results_0.update({source_name: {}})
for target_name, target_conf in files.iteritems():
if not source_name == target_name:
r_diff = []
for _ in xrange(n_reps):
"Construct a prediction model using source"
predict_model = train_prediction_model(source_conf,
T=5)
"""Sample 15 from train and test datasets
to train a transfer model
"""
"Find common configs between source and target"
common = pd.merge(source_conf,
target_conf,
how="inner")
"Pick random 15 samples"
some = common.sample(n=10)
"Get the dependent variables to construct a LR model"
p_src = some[source_conf.columns[-1]]
p_tgt = some[target_conf.columns[-1]]
"Train a transfer model"
transfer_model = train_transfer_model(p_src=p_src,
p_tgt=p_tgt)
"Remove elements used to train transfer model from target"
target_conf = target_conf.drop(some.index,
errors="ignore")
"Perform tansfer"
target_indep = target_conf[target_conf.columns[:-1]]
target_actual = target_conf[target_conf.columns[-1]]
predicted_raw = predict_model.predict(
target_indep).reshape(-1, 1)
target_predicted = transfer_model.predict(
predicted_raw).reshape(1, -1)[0]
"Get rank difference"
r_diff.append(
rank_diff(actual=target_actual,
predicted=target_predicted))
results_0[source_name].update(
{target_name: int(np.median(r_diff))})
results.update({project.name: pd.DataFrame(results_0)})
# -------------------- DEBUG -------------------- #
set_trace()
def compare(self, data_pairs):
data_path = os.path.realpath("./data")
projects = get_all_projects(data_path)
for project in projects:
print(project.name.upper())
files = project.files()
best, worst = data_pairs[project.name]
rest = [
dframe for fname, dframe in files.iteritems()
if fname not in data_pairs[project.name]
]
best_results = ["best"]
worst_results = ["worst"]
for tgt in rest:
best_results.extend(
[Pooyan.learner(files[best], tgt) for _ in xrange(1)])
worst_results.extend(
[Pooyan.learner(files[worst], tgt) for _ in xrange(1)])
rdivDemo([best_results, worst_results])
set_trace()
def main(n_reps=30):
data_path = os.path.realpath("./data")
projects = get_all_projects(data_path)
results = dict()
for project in projects:
files = project.files()
results_0 = dict()
for source_name, source_conf in files.iteritems():
results_0.update({source_name: {}})
for target_name, target_conf in files.iteritems():
if not source_name == target_name:
r_diff = []
for _ in xrange(n_reps):
"Get the dependent variables to construct a LR model"
p_src = source_conf[source_conf.columns[-1]]
p_tgt = target_conf[target_conf.columns[-1]]
"Construct a gaussian process model using source"
predict_model = train_gaussproc_model(
source_conf, target_conf)
"Perform tansfer"
target_indep = target_conf[target_conf.columns[:-1]]
target_actual = target_conf[target_conf.columns[-1]]
target_predicted = predict_model.predict(
target_indep).reshape(-1, 1)
"Get rank difference"
r_diff.append(
rank_diff(actual=target_actual,
predicted=target_predicted))
results_0[source_name].update(
{target_name: int(np.median(r_diff))})
results.update({project.name: pd.DataFrame(results_0)})
# -------------------- DEBUG -------------------- #
set_trace()