def test_data_size_vs_diff(dm, given_dict, infer_dict):
#Read all data from data model
dm.read_data(normalize_data=False)
#attr_list = [U_UNIVERSITY_CODE, PROGRAM_CODE, UNIVERSITY, MAJOR_CODE, TERM]
attr_list = [U_UNIVERSITY_CODE, PROGRAM_CODE, UNIVERSITY]
#attr_list = [MAJOR_CODE, PROGRAM_CODE, TERM]
#Size of data
data_size = len(dm.data)
#Step size = 10 steps
step_size = data_size//10
#Get experiment data in a dict
size = []
accuracy = []
for i in xrange(step_size, data_size, step_size):
dm_test = DataModel("")
dm_test.set_data(dm.data[:i])
exp_test = Experimenter(dm_test, attr_list)
actual = exp_test.get_actual_result(given_dict, infer_dict)
estimation = exp_test.generic_get_estimated_result(given_dict, infer_dict)
size.append(i)
accuracy.append(abs(estimation - actual))
print("Step:%d--->Actual:%f--->Estimate:%f" %(i, actual, estimation))
print "-------------------------------------------------------------"
plt.figure()
plt.plot(size, accuracy)
plt.title("Data Size vs Accuracy")
plt.show()
def main(args):
dm = DataModel(args.gig_file, args.chat_file)
dm.read_data()
exp = Experimenter(dm)
if args.classify is True:
scores = exp.classify_gigs()
if args.feature_values is True:
scores = exp.evaluate_feature_values()
return dm
def main():
debug = 0
sample_size = 8192
#database = MFPT(debug=debug)
database = Paderborn(debug=debug)
database_acq = database.load()
#print(database_acq)
database_exp = Experimenter(database_acq, sample_size)
database_exp.perform(Classifiers(), Scoring())
def __init__(self, trials=3):
self.exp = Experimenter()
self.mongo_coll_conn = Mongo(collection_name='optimisation')
self.trials = trials
self.config = load_config()
self.task_type = self.config['Utils']['task_type']
self.data_name = self.config['Utils'][self.task_type]['data_name']
def main(args):
dm = DataModel(args.data_file)
dm.read_data(to_read_count=10000)
exp = Experimenter(dm, \
process_datamodel=True, \
serialise=False)
t1 = time.time()
exp.perform_multiclass_experiment(
pred_mode=INDEPENDENT,
use_exclusion=True,
need_to_extract_features=True,
prediction_file='../results/predictions_multiclass_independent_englishonly_legibleonly_wordunibigram_chartrigram_10000.csv',
result_file='../results/results_multiclass_independent_englishonly_legibleonly_wordunibigram_chartrigram_10000.txt',
english_only=True,
legible_only=True)
t2 = time.time()
timeused = t2 - t1
logging.getLogger(LOGGER).info('Time used in experiment (hour:min:sec): %d:%d:%d' % \
(timeused/3600, timeused/60, timeused%60))
return exp
def main(args):
dm = DataModel(args.data_file)
dm.read_data(to_read_count=10000)
exp = Experimenter(dm, \
process_datamodel=True, \
serialise=False)
t1 = time.time()
exp.perform_multiclass_experiment(
pred_mode=INDEPENDENT,
use_exclusion=True,
need_to_extract_features=True,
prediction_file=
'../results/predictions_multiclass_independent_englishonly_legibleonly_wordunibigram_chartrigram_10000.csv',
result_file=
'../results/results_multiclass_independent_englishonly_legibleonly_wordunibigram_chartrigram_10000.txt',
english_only=True,
legible_only=True)
t2 = time.time()
timeused = t2 - t1
logging.getLogger(LOGGER).info('Time used in experiment (hour:min:sec): %d:%d:%d' % \
(timeused/3600, timeused/60, timeused%60))
return exp
def main(args):
dm = DataModel(args.train_file)
dm.read_train_data()
exp = Experimenter(dm)
distances = [x.get_distance() for x in dm.data]
print(max(distances))
print(min(distances))
print(stats.mean(distances))
t1 = time.time()
t2 = time.time()
timeused = t2 - t1
logging.getLogger(LOGGER).info('Time used in experiment (hour:min:sec): %d:%d:%d' % \
(timeused/3600, timeused/60, timeused%60))
return exp
if args.check_estimator:
common_params['eval_metrics'] += \
'-CPrecIPSin0.0_10-CPrecIPSin0.0_100-CDCGIPSin0.0_100000-CARIPSin0.0' + \
'-CPrecIPSin0.001_10-CPrecIPSin0.001_100-CDCGIPSin0.001_100000-CARIPSin0.001' + \
'-CPrecIPSin0.003_10-CPrecIPSin0.003_100-CDCGIPSin0.003_100000-CARIPSin0.003' + \
'-CPrecIPSin0.01_10-CPrecIPSin0.01_100-CDCGIPSin0.01_100000-CARIPSin0.01' + \
'-CPrecIPSin0.03_10-CPrecIPSin0.03_100-CDCGIPSin0.03_100000-CARIPSin0.03' + \
'-CPrecIPSin0.1_10-CPrecIPSin0.1_100-CDCGIPSin0.1_100000-CARIPSin0.1' + \
'-CPrecIPSin0.3_10-CPrecIPSin0.3_100-CDCGIPSin0.3_100000-CARIPSin0.3'
else:
common_params['eval_metrics'] = 'Prec_10-Prec_100' + '-CPrec_10-CPrec_100-CDCG_100000-CAR'
if args.check_estimator:
common_params['eval_metrics'] += '-CPrecIPSin0.01_10-CPrecIPSin0.01_100-CDCGIPSin0.01_100000-CARIPSin0.01'
# set up experimenter
experimenter = Experimenter()
list_params = experimenter.set_search_params(args.cond_search, args.type_search)
list_params = experimenter.set_common_params(list_params, common_params)
save_result_file = dir_data_prepared + "result/" + datetime.now().strftime(
'%Y%m%d_%H%M%S') + "_" + args.type_model + "_" + args.name_experiment + '_tlt' + str(args.time_length_train) + ".csv"
if phase == 'test_phase':
save_result_file = save_result_file.replace('.csv', '_test.csv')
print('save_result_file is {}'.format(save_result_file))
save_result_dir = os.path.dirname(save_result_file)
if not os.path.exists(save_result_dir):
os.mkdir(save_result_dir)
print('Start experiment.')
t_init = datetime.now()
"num_factor": 40
}
files["movietweetings-gte.csv"] = {
"learning_rate": 0.0001,
"reg_rate": 0.0,
"batch_size": 256,
"num_factor": 40
}
files["ml-100k-gte.csv"] = {
"learning_rate": 0.0005,
"reg_rate": 0.0,
"batch_size": 256,
"num_factor": 40
}
experimenter = Experimenter()
experimenter.config_gpu()
experimenter.addSamplingApproach(Cosine)
experimenter.addSamplingApproach(TF_IDF)
experimenter.addSamplingApproach(ARM)
experimenter.addSamplingApproach(Random)
experimenter.addMaxRejection(TotalLimit)
experimenter.addMaxRejection(UniqueLimit)
experimenter.addMaxRejection(Q3Total)
experimenter.addMaxRejection(Q3Unique)
experimenter.setModel(NeuMF)
experimenter.setParameterFiles(files)
experimenter.execute()
from experiments import EXPERIMENTS
from experimenter import Experimenter
import utils
parser = argparse.ArgumentParser()
parser.add_argument('-n',
'--trials',
type=int,
default=1,
help="How many trials?")
parser.add_argument('-tb',
'--tensorboard',
action='store_true',
help="Should we write to TensorBoard")
rng = np.random.default_rng()
experiments = [0]
# experiments = list(sorted(EXPERIMENTS.keys()))
# experiments = list(range(31, 51))
# Experiment 0 is a test experiment
# experiments.remove(0)
if __name__ == '__main__':
args = parser.parse_args()
print(f"Running {args.trials} trial(s) of experiments {experiments}.")
exp_runner = Experimenter(write_to_tensorboard=args.tensorboard)
for exp in experiments:
for _ in range(args.trials):
exp_runner.run_experiment(exp, rng)
