def test_models():
for i, model in enumerate(models):
print("model{0} results:".format(i + 1))
# dohvati vjerojatnosti na skupu za učenje
probs = model.eval(mnist.train.images)
# ispiši performansu (preciznost i odziv po razredima)
Y = model.eval_class(mnist.test.images)
accuracy, pr, M = data.eval_perf_multi(
Y, mnist.test.labels.argmax(axis=1))
print("Accuracy: ", accuracy)
print("Precision / Recall: ", pr)
print("Confusion Matrix: ", M)
tf.set_random_seed(100)
# instanciraj podatke X i labele Yoh_
X, Y_ = data.sample_gmm_2d(6, 2, 10)
Yoh_ = data.class_to_onehot(Y_)
# izgradi graf:
tf_deep = TF_deep([2, 10, 10, 2])
# nauči parametre:
tf_deep.train(X, Yoh_, 10000)
# dohvati vjerojatnosti na skupu za učenje
probs = tf_deep.eval(X)
print(probs)
# ispiši performansu (preciznost i odziv po razredima)
Y = tf_deep.eval_class(X)
print(Y)
accuracy, pr, M = data.eval_perf_multi(tf_deep.eval_class(X), Y_)
print("Accuracy: ", accuracy)
print("Precision / Recall: ", pr)
print("Confusion Matrix: ", M)
# iscrtaj rezultate, decizijsku plohu
rect = (np.min(X, axis=0), np.max(X, axis=0))
data.graph_surface(tf_deep.eval_probs, rect, offset=0.5)
data.graph_data(X, Y_, Y_)
plt.show()
def fcann2_classify(X, W1, b1, W2, b2):
s1 = np.dot(X, W1) + b1
h1 = relu(s1)
s2 = np.dot(h1, W2) + b2
return softmax(s2)
if __name__ == "__main__":
# init random seeds
np.random.seed(100)
# create data
X, Y_ = data.sample_gmm_2d(6, 2, 10)
# train model
W1, b1, W2, b2 = fcann2_train(X, Y_, 5, 10000, 0.05, 1e-3, print_diagnostics_step=1000)
# print diagnostics
Y = fcnn2_decfun_class(X, W1, b1, W2, b2)(X)
accuracy, pr, M = data.eval_perf_multi(Y, Y_)
print("Accuracy: ", accuracy)
print("Precision / Recall: ", pr)
print("Confusion Matrix: ", M)
# graph the decision surface
rect = (np.min(X, axis=0), np.max(X, axis=0))
decision_function = fcnn2_decfun_probs(X, W1, b1, W2, b2)
data.graph_surface(decision_function, rect, offset=0.5)
data.graph_data(X, Y_, fcnn2_decfun_class(X, W1, b1, W2, b2))
plt.show()
def get_scores(self, X, Y_):
accuracy, pr, M = data.eval_perf_multi(self.predict(X), Y_)
print("Accuracy: ", accuracy)
print("Precision / Recall: ", pr)
print("Confusion Matrix: ", M)