q1.py

import numpy as np 
import matplotlib.pyplot as plt
from svm import SVM

np.random.seed(1)

def get_data(lower,upper,num,num_dims):
	return np.random.uniform(lower,upper,size=(num,num_dims))

def get_labels(X):
	Y = []
	for x1,x2 in X:
		if x2 < np.sin(10*x1)/5 + 0.3 or ((x2 - 0.8)**2 + (x1 - 0.5)**2)<0.15**2:
			Y.append(1)
		else:
			Y.append(-1)
	return np.asarray(Y)

def main():
	N = 100
	data = get_data(0,1,N,2) 
	labels = get_labels(data).reshape(-1)
	predictions = np.ones_like(labels)*-1
	print("Max-class classifier training set accuracy: ",np.mean(np.equal(predictions,labels))*100,"%")
	model = SVM(kernel="rbf",gamma=3)
	model.fit(data,labels)
	predictions = model.predict(data)
	print("SVM model Training set accuracy: ",np.mean(np.equal(predictions,labels))*100,"%")
	print("Number of SVMs computed: ",model._n_support)
	print("Value of intercept: ",model.intercept)
	# print(model.project(data))

	color = np.where(model._support_labels==1,"orange","green")
	plt.scatter(data[:, 0], data[:, 1], c=labels, s=50, cmap=plt.cm.bwr)
	plt.scatter(model._support_vectors[:, 0], model._support_vectors[:, 1], s=35, c=color, marker='H')
	plt.title('SVM Boundaries (N = %d)' % (N))

	X1, X2 = np.meshgrid(np.linspace(0, 1, 100), np.linspace(0, 1, 100))
	X_T = np.array([[x1, x2] for x1, x2 in zip(np.ravel(X1), np.ravel(X2))])

	Z = model.project(X_T).reshape(X1.shape)

	H = plt.contour(X1, X2, Z, [0.0], colors='k', linewidths=2, origin='lower')
	H_1= plt.contour(X1, X2, Z + 1, [0.0],colors='tab:green', linewidths=1, origin='lower')
	H1 = plt.contour(X1, X2, Z - 1, [0.0], colors='orange', linewidths=1, origin='lower')

	plt.clabel(H,inline=True, fmt="0", fontsize=8)
	plt.clabel(H_1,inline=True, fmt="-1", fontsize=8)
	plt.clabel(H1,inline=True, fmt="+1", fontsize=8)
	plt.axis("tight")
	plt.show()
	
if __name__ == '__main__':
	main()