error += 1 elif predict_c != Y[i]: error += 1 return float(error) / X.shape[0] #Parameters here c = 1 kernal = linear_kernel #Parameters here alpha, theta, theta_0 = multi_svm_train(X_train, Y_train, c, kernal) scoreFns = [] for c in alpha.keys(): score = ScoreFns(X_train, Y_train, kernal, alpha[c], theta_0[c]) scoreFns.append(score.fn) plotMultiDecisionBoundary(X_train, Y_train, scoreFns, [0, 0, 0], title="Linear Kernal, toy_1") #c = 1 #alpha, theta, theta_0 = multi_svm_train(X_train, Y_train, c, poly_kernel) #er = error_rate(X_val, Y_val, X_train, Y_train, alpha, theta_0, poly_kernel) #print er """ classes = np.sort(np.unique(Y_train), axis=None) for c in classes: Y_plot = np.array(Y_train == c, dtype=float) # define your matrices Y_lot = Y_plot * 2 - 1.0 #for c in [1e3]:
return float(error) / X.shape[0] """ #Parameters here c = 100 kernel = poly_kernel #Parameters here alpha, theta, theta_0 = multi_svm_train(X_train, Y_train, c, kernel) er_train = error_rate(X_train, Y_train, X_train, Y_train, alpha, theta_0, kernel) print "er_train is {}".format(er_train) er_val = error_rate(X_val, Y_val, X_train, Y_train, alpha, theta_0, kernel) print "er_val is {}".format(er_val) er_test = error_rate(X_test, Y_test, X_train, Y_train, alpha, theta_0, kernel) print "er_test is {}".format(er_test) """ """ scoreFns = [] for c in alpha.keys(): score = ScoreFns(X_train, Y_train, kernal, alpha[c], theta_0[c]) scoreFns.append(score.fn) plotMultiDecisionBoundary(X_train, Y_train, scoreFns, [0, 0, 0], title="Linear Kernal, toy_1") """ #c = 1 #alpha, theta, theta_0 = multi_svm_train(X_train, Y_train, c, poly_kernel) #er = error_rate(X_val, Y_val, X_train, Y_train, alpha, theta_0, poly_kernel) #print er """ #for c in [1e3]: