def experiment(Xtrn, Ctrn, Xtst, Ctst):
classes = np.unique(Ctrn)
N = np.size(classes)
Kns_vector = []
for i in classes:
acc_list = []
labels_list = []
for j in NUM_OF_CLUSTERS_TO_TEST:
labels = my_kmeans_classify(get_class_vectors(Xtrn, Ctrn, i), j)
new_Ctrn = update_labels(Ctrn, labels, i, N)
[Cpreds, _, _] = my_gaussian_classify(Xtrn, new_Ctrn, Xtst)
preds = normalize_preds(Cpreds, i, N)
[_, acc] = conf(Ctst, preds)
acc_list.append(acc)
max = np.argmax(acc_list)
Kns_vector.append(NUM_OF_CLUSTERS_TO_TEST[max])
return np.array(Kns_vector).transpose()
from data_utils import load_data
from my_knn_classify import my_knn_classify as knn
from my_confusion import my_confusion as conf
import time
import numpy as np
import scipy.io as sc
Ks = [1, 3, 5, 10, 20]
[Xtrn, Ctrn, Xtst, Ctst] = load_data()
start = time.clock()
Cpreds = knn(Xtrn, Ctrn, Xtst, Ks)
end = time.clock()
print "Time taken: ", (end - start)
N = np.size(Ctst)
for i in range(np.size(Ks)):
[CM, acc] = conf(Ctst, Cpreds[:, i].reshape(N, 1))
print "Number of nearest neighbours: ", Ks[i]
print "Number of samples: ", N
print "Number of misclassifications: ", int(N - np.trace(CM))
print "Accuracy: ", acc
print "--------------------------------"
filename = "cm" + str(Ks[i])
sc.savemat(filename, {"Confusion_matrix": CM})
from data_utils import load_data
from my_gaussian_classify import my_gaussian_classify as gaussian
from my_confusion import my_confusion as conf
import numpy as np
import time
import scipy.io as sc
[Xtrn, Ctrn, Xtst, Ctst] = load_data()
start = time.clock()
[Cpreds, Ms, Covs] = gaussian(Xtrn, Ctrn, Xtst, 0.01)
end = time.clock()
print "Time taken: ", (end - start)
[cm, acc] = conf(Ctst, Cpreds)
sc.savemat("cm", {"Confusion_matrix": cm})
sc.savemat("m26", {"Mean_vector": Ms[:, 25]})
sc.savemat("cov26", {"Covariance_matrix": Covs[:, :, 25]})
N = np.size(Ctst)
print "Number of samples: ", N
print "Number of misclassifications: ", int(N - np.trace(cm))
print "Accuracy: ", acc