from scipy.io import loadmat
import matplotlib.pyplot as plt
import numpy as np
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score
from nfold import nFold
mat = loadmat('Assignment_2_datasets/dataset_2.mat')
samples = mat['samples']
labels = mat['labels'][0]
train_samples,train_labels,test_samples,test_labels = nFold(samples,labels,5,4)
clf = LogisticRegression(random_state=0).fit(samples, labels)
pred = clf.predict(samples)
print(accuracy_score(pred,labels))
print(clf.coef_)
print(clf.intercept_)
def mse(pred, test_l):
k = pred - test_l
k = np.multiply(k, k)
return np.mean(k)
samples, labels = load_data()
fold = []
tr_err = []
te_err = []
sk_tr_err = []
sk_te_err = []
for i in range(5): #FOR 5 FOLDS
train_s, train_l, test_s, test_l = nFold(samples, labels, 5, i)
model = Regression()
model.fit(train_s, train_l)
train_pred = model.predict(train_s)
test_pred = model.predict(test_s)
fold.append(i)
tr_err.append(mse(train_pred, train_l))
te_err.append(mse(test_pred, test_l))
sk_tr_err.append(mean_squared_error(train_pred, train_l))
sk_te_err.append(mean_squared_error(test_pred, test_l))
with open('Q1_b.csv', 'w') as file:
writer = csv.writer(file)
writer.writerow([
plt.plot(lam_list, acc)
plt.title('Grid Search')
plt.xlabel('Lambda')
plt.ylabel('Accuracy')
plt.show()
return lam_list[np.argmax(acc)]
mat = loadmat('Assignment_2_datasets/dataset_1.mat')
samples = mat['samples']
labels = mat['labels'][0]
samples = np.append(samples, np.ones((len(labels), 1)),
axis=1) #Adding a column of 1 for constant
train_s, train_l, test_s, test_l = nFold(samples, labels, 5,
4) #Grid search for lambda value
#lam = perform_grid_search(train_s,train_l,test_s,test_l)
lam = 5
tr_acc = []
te_acc = []
tr_loss = []
te_loss = []
for i in range(5):
train_s, train_l, test_s, test_l = nFold(samples, labels, 5, i)
model = LogRegression(lam)
iter, train_acc, train_loss, test_acc, test_loss = model.fit(
train_s, train_l, test_s, test_l)