def DrawRBFContour(centers, w_lin, data, xmin, xmax, ymin, ymax):
start_time = time.time()
xlist = np.linspace(xmin, xmax, 250)
ylist = np.linspace(ymin, ymax, 250)
X, Y = np.meshgrid(xlist, ylist)
Z = []
for x_row, y_row in zip(X, Y):
z_row = zip(x_row, y_row)
Z.append(z_row)
new_Z = []
k = len(centers)
r = 2 / (np.sqrt(k))
for z_row in Z:
row = [
RBF_Classify(w_lin, centers, test_point, r) for test_point in z_row
]
new_Z.append(row)
#plt.figure()
cp = plt.contourf(X, Y, new_Z)
plt.colorbar(cp)
plt.title('RBF Plot')
plt.xlabel('X1')
plt.ylabel('X2')
cls.draw_data(data)
print("--- %s seconds ---" % (time.time() - start_time))
plt.show()
def DrawNNContour(k, data, trans, xmin, xmax, ymin, ymax, use_sort):
start_time = time.time()
xlist = np.linspace(xmin, xmax, 250)
ylist = np.linspace(ymin, ymax, 250)
X, Y = np.meshgrid(xlist, ylist)
Z=[]
for x_row, y_row in zip(X,Y):
z_row=zip(x_row,y_row)
Z.append(z_row)
new_Z=[]
if trans:
trans_data = [(Transform(x),y) for x,y in data]
for z_row in Z:
row=[KNearestNeighbor(k, trans_data, Transform(test_point), use_sort) for test_point in z_row]
new_Z.append(row)
else:
for z_row in Z:
row=[KNearestNeighbor(k, data, test_point, use_sort) for test_point in z_row]
new_Z.append(row)
#plt.figure()
cp = plt.contourf(X, Y, new_Z)
plt.colorbar(cp)
plt.title('NN Plot')
plt.xlabel('X1')
plt.ylabel('X2')
cls.draw_data(data)
print("--- %s seconds ---" % (time.time() - start_time))
plt.show()
def DrawGGContour(data, xmin, xmax, ymin, ymax):
xlist = np.linspace(xmin, xmax, 100)
ylist = xlist**3
plt.plot(xlist, ylist, "g-")
x_1 = [0, 0.00001]
y_1 = [ymin, ymax]
plt.xlim(-1.2, 1.2)
plt.ylim(-1.2, 1.2)
plt.plot(x_1, y_1, "b-")
cls.draw_data(data)
plt.show()
def DrawLinearContour(data, xmin, xmax, ymin, ymax):
start_time = time.time()
xlist = np.linspace(xmin, xmax, 250)
ylist = np.linspace(ymin, ymax, 250)
X, Y = np.meshgrid(xlist, ylist)
Z=np.sign(X+2*Y+2.23)
#plt.figure()
cp = plt.contourf(X, Y, Z)
plt.colorbar(cp)
plt.title('NeuralNet Plot')
plt.xlabel('X1')
plt.ylabel('X2')
cls.draw_data(data)
print("--- %s seconds ---" % (time.time() - start_time))
plt.show()
def DrawLegendreContour(order, g, train_data, test_data):
##Draw contour==============================
print "Drawing Contour..."
x1_list = np.arange(-1.2, 1.2, 0.01)
x2_list = np.arange(-1.2, 1.2, 0.01)
X1, X2 = np.meshgrid(x1_list, x2_list)
Z = []
transformed = [LegendreTransform(order, x1, x2) for x1, x2 in zip(X1, X2)]
for x in transformed:
weighted = sum([coef * feature for coef, feature in zip(g, x)])
Z.append(weighted)
##=============================================
cls.draw_data(train_data)
plt.contour(X1, X2, Z, [0])
plt.axis([-1.1, 1.1, -1.1, 1.1])
plt.show()
cls.draw_data(test_data)
plt.contour(X1, X2, Z, [0])
plt.axis([-1.1, 1.1, -1.1, 1.1])
plt.show()
max_iteration = 10000
accuracy_threshold = 0.999
##================================================
## Data preprocessing=====================================
training_data = extractDigits("ZipDigits.train")
test_data = extractDigits("ZipDigits.test")
raw_train_labeled_data = labelImageData(training_data, d1, d2,
features.horiAsymmetry,
features.vertAsymmetry)
raw_test_labeled_data = labelImageData(test_data, d1, d2,
features.horiAsymmetry,
features.vertAsymmetry)
if see_train: cls.draw_data(raw_train_labeled_data)
else: cls.draw_data(raw_test_labeled_data)
if not tran:
trans_train_data = raw_train_labeled_data
trans_test_data = raw_test_labeled_data
else:
trans_train_data = thirdOrderTransfrom(raw_train_labeled_data)
trans_test_data = thirdOrderTransfrom(raw_test_labeled_data)
##===========================================================
##Learning=========================================================
g = regressionAndPocket(trans_train_data, accuracy_threshold,
max_iteration)
##===========================================================
