def svm2():
train_data = np.load('hw3_data/nonlinsep/train_data.npy')
train_labels = np.load('hw3_data/nonlinsep/train_labels.npy')
k_list = ['linear', 'rbf', 'poly', 'sigmoid']
for k in k_list:
clf = svm.SVC(kernel=k)
clf.fit(train_data, train_labels)
draw_svm(clf, train_data, train_labels, -3, 3, -3, 3, 'svm1_' + k)
def svm1():
train_data = np.load('hw3_data/linsep/train_data.npy')
train_labels = np.load('hw3_data/linsep/train_labels.npy')
c_list = [0.01, 0.1, 1, 10, 100]
for i in range(len(c_list)):
clf = svm.SVC(kernel='linear', C=c_list[i])
clf.fit(train_data, train_labels)
draw_svm(clf, train_data, train_labels, -3, 3, -3, 3, 'svm1_' + str(i))
def svm_part_1(c: float, data: np.array, labels: np.array, save: bool = True):
model = SVC(C=c, kernel='linear')
model.fit(data, labels)
x_values = [x[0] for x in data]
y_values = [x[1] for x in data]
min_x, min_y = min(x_values), min(y_values)
max_x, max_y = max(x_values), max(y_values)
if save is True:
draw_svm(model, data, labels, min_x, max_x, min_y, max_y, f"plots/svm/q1-{c}.png")
else:
draw_svm(model, data, labels, min_x, max_x, min_y, max_y)
def task2():
kernel = sys.argv[2]
train_data = np.load('hw3_data/nonlinsep/train_data.npy')
train_labels = np.load('hw3_data/nonlinsep/train_labels.npy')
cf = SVC(kernel=kernel)
cf.fit(train_data, train_labels)
x_min, x_max, y_min, y_max = min(train_data[:, 0]), max(
train_data[:, 0]), min(train_data[:, 1]), max(train_data[:, 1])
draw.draw_svm(cf, train_data, train_labels, x_min, x_max, y_min, y_max)
def task1():
c = float(sys.argv[2])
train_data = np.load('hw3_data/linsep/train_data.npy')
train_labels = np.load('hw3_data/linsep/train_labels.npy')
cf = SVC(c, kernel='linear')
cf.fit(train_data, train_labels)
x_min, x_max, y_min, y_max = min(train_data[:, 0]), max(
train_data[:, 0]), min(train_data[:, 1]), max(train_data[:, 1])
draw.draw_svm(cf, train_data, train_labels, x_min, x_max, y_min, y_max)
def svm_part_2(kernel: str,
data: np.array,
labels: np.array,
save: bool = True):
model = SVC(C=1, kernel=kernel)
model.fit(data, labels)
x_values = [x[0] for x in data]
y_values = [x[1] for x in data]
min_x, min_y = min(x_values), min(y_values)
max_x, max_y = max(x_values), max(y_values)
if save is True:
draw_svm(model, data, labels, min_x, max_x, min_y, max_y,
f"plots/svm/q2-{kernel}.png")
else:
draw_svm(model, data, labels, min_x, max_x, min_y, max_y)
def part1():
train_data = np.load('hw3_data/linsep/train_data.npy')
train_labels = np.load('hw3_data/linsep/train_labels.npy')
c_values = [0.01, 0.1, 1, 10, 100]
for c in c_values:
clf = SVC(C=c, kernel='linear')
clf.fit(train_data, train_labels)
draw_svm(clf,
train_data,
train_labels,
-3,
3,
-3,
3,
target='part1_' + str(c) + ".png")
def part2():
train_data = np.load('hw3_data/nonlinsep/train_data.npy')
train_labels = np.load('hw3_data/nonlinsep/train_labels.npy')
kernels = ["linear", "rbf", "poly", "sigmoid"]
for k in kernels:
clf = SVC(C=1, kernel=k)
clf.fit(train_data, train_labels)
draw_svm(clf,
train_data,
train_labels,
-3,
3,
-3,
3,
target='part2_' + str(k) + ".png")