test_theorems = set(theorems) - set(train_theorems)
params_data_trans = {'features': features, 'chronology': chronology}
train_labels, train_array = prs.proofs_to_train(proofs_train,
params_data_trans,
n_jobs=N_JOBS,
logfile=LOG_FILE)
params_train = {}
model = prs.train(train_labels,
train_array,
params=params_train,
n_jobs=N_JOBS,
logfile=LOG_FILE)
rankings_train = prs.Rankings(train_theorems,
model,
params_data_trans,
n_jobs=N_JOBS,
logfile=LOG_FILE)
rankings_test = prs.Rankings(test_theorems,
model,
params_data_trans,
n_jobs=N_JOBS,
logfile=LOG_FILE)
params_atp_eval = {}
proofs_train.update(
prs.atp_evaluation(rankings_train,
statements,
params_atp_eval,
dirpath=ATP_DIR,
n_jobs=N_JOBS,
logfile=LOG_FILE))
statements = prs.Statements(from_file=join(DATA_DIR, 'statements'),
logfile=LOG_FILE)
features = prs.Features(from_file=join(DATA_DIR, 'features'), logfile=LOG_FILE)
chronology = prs.Chronology(from_file=join(DATA_DIR, 'chronology'),
logfile=LOG_FILE)
theorems = prs.utils.readlines(join(DATA_DIR, 'theorems_atpproved'))
params_data_trans = {
'features': features,
'chronology': chronology,
'only_short_proofs': False
}
# randomly generated rankings
rankings_random = prs.Rankings(theorems,
model=None,
params=params_data_trans,
n_jobs=N_JOBS,
logfile=LOG_FILE)
proofs = prs.atp_evaluation(rankings_random,
statements,
dirpath=ATP_DIR,
n_jobs=N_JOBS,
logfile=LOG_FILE)
for i in range(40):
prs.utils.printline("ITERATION: {}".format(i), LOG_FILE)
train_labels, train_array = prs.proofs_to_train(proofs,
params_data_trans,
n_jobs=N_JOBS,
logfile=LOG_FILE)