def finish_part_c(train_X,
train_y,
test_X,
test_y,
data_type,
seed_value,
verbose=True,
to_return=False):
"""
Finishes the part(c) of the homework5 of EE559.
:param train_X: feature matrix for training data.
:param train_y: label vector for training data.
:param test_X: feature matrix for test data.
:param test_y: label vector for test data.
:param data_type: "unnormalized" or "standardized".
:param seed_value: value to control starting weight vectors.
:param verbose: whether to print information.
:param to_return: whether to return accuracy value or not.
Outputs the requried information.
"""
perceptron_model = Perceptron(max_iter=10000, tol=None, random_state=0)
np.random.seed(seed_value)
coef = np.random.rand(len(np.unique(train_y)), train_X.shape[1])
intercept = np.random.rand(len(np.unique(train_y)), 1)
perceptron_OvR = OneVsRestClassifier(estimator=perceptron_model)
perceptron_OvR.coef_ = coef # Initialize coefficients with the specific seed value
perceptron_OvR.intercept_ = intercept # Initialize intercepts with the specific seed value
perceptron_OvR.fit(train_X, train_y)
train_y_pred = perceptron_OvR.predict(train_X)
test_y_pred = perceptron_OvR.predict(test_X)
if verbose:
print("The weights for each feature for " + data_type + " data",
perceptron_OvR.coef_)
print("The constants for each feature for " + data_type + " data",
perceptron_OvR.intercept_)
if to_return:
return get_accuracy(train_y,
train_y_pred,
log_word=data_type + " training data",
to_return=to_return)
else:
get_accuracy(train_y,
train_y_pred,
log_word=data_type + " training data")
get_accuracy(test_y, test_y_pred, log_word=data_type + " test data")
print(" ")
