def _train_svr(mat_files, out_file, C=1000, gamma=0.001, epsilon=0.01):
"""
Trains SVR with the given parameters
"""
peptides, signals = _get_peptides_signals(mat_files)
model = SvrBlockade()
model.train(peptides, signals, C, gamma, epsilon)
store_model(model, out_file)
def _train_random_forest(mat_files, out_file):
"""
Trains Random Forest
"""
peptides, signals = _get_peptides_signals(mat_files)
model = RandomForestBlockade()
model.train(peptides, signals)
store_model(model, out_file)
def _train_svr(mat_files, out_file, C=1000, gamma=0.001, epsilon=0.01):
"""
Trains SVR with the given parameters
"""
peptides, signals = _get_peptides_signals(mat_files)
model = SvrBlockade()
model.train(peptides, signals, C, gamma, epsilon)
store_model(model, out_file)
def _train_random_forest(mat_files, out_file):
"""
Trains Random Forest
"""
peptides, signals = _get_peptides_signals(mat_files)
model = RandomForestBlockade()
model.train(peptides, signals)
store_model(model, out_file)
def _cross_validate(train_mats, cv_mats, db_file, out_file):
"""
Choosing the best parameters through cross-validation
"""
CLUSTER_SIZE = 10
eps_vec = [0.01, 0.001, 0.0001, 0.00001]
C_vec = [1, 10, 100, 1000, 10000, 100000]
gamma_vec = [0.00001, 0.0001, 0.001, 0.01, 0.1, 1]
best_score = sys.maxint
best_params = None
print("C\tGam\tEps\tScore", file=sys.stderr)
for C in C_vec:
for gamma in gamma_vec:
for eps in eps_vec:
temp_model = _train_svr(train_mats, C, gamma, eps)
scores = []
for cv_mat in cv_mats:
pval, rank = pvalues_test(cv_mat, CLUSTER_SIZE, temp_model,
db_file, False,
open(os.devnull, "w"))
scores.append(rank)
score = np.mean(scores)
print("{0}\t{1}\t{2}\t{3}".format(C, gamma, eps, score),
file=sys.stderr)
if score < best_score:
best_score = score
best_params = (C, gamma, eps)
print(*best_params, file=sys.stderr)
best_model = _train_svr(train_mats, *best_params)
store_model(best_model, out_file)
def _cross_validate(train_mats, cv_mats, db_file, out_file):
"""
Choosing the best parameters through cross-validation
"""
CLUSTER_SIZE = 10
eps_vec = [0.01, 0.001, 0.0001, 0.00001]
C_vec = [1, 10, 100, 1000, 10000, 100000]
gamma_vec = [0.00001, 0.0001, 0.001, 0.01, 0.1, 1]
best_score = sys.maxint
best_params = None
print("C\tGam\tEps\tScore", file=sys.stderr)
for C in C_vec:
for gamma in gamma_vec:
for eps in eps_vec:
temp_model = _train_svr(train_mats, C, gamma, eps)
scores = []
for cv_mat in cv_mats:
pval, rank = pvalues_test(cv_mat, CLUSTER_SIZE, temp_model,
db_file, False,
open(os.devnull, "w"))
scores.append(rank)
score = np.mean(scores)
print("{0}\t{1}\t{2}\t{3}".format(C, gamma, eps, score),
file=sys.stderr)
if score < best_score:
best_score = score
best_params = (C, gamma, eps)
print(*best_params, file=sys.stderr)
best_model = _train_svr(train_mats, *best_params)
store_model(best_model, out_file)
