def deserialize_gaussian_nb(model_dict):
model = GaussianNB(model_dict['params'])
model.classes_ = np.array(model_dict['classes_'])
model.class_count_ = np.array(model_dict['class_count_'])
model.class_prior_ = np.array(model_dict['class_prior_'])
model.theta_ = np.array(model_dict['theta_'])
model.sigma_ = np.array(model_dict['sigma_'])
model.epsilon_ = model_dict['epsilon_']
return model
def NB_predict():
X = json.loads(request.form['X'])
params = json.loads(request.form['params'])
clf = GaussianNB()
clf.class_prior_ = np.array(params['class_prior'])
clf.class_count_ = np.array(params['class_count'])
clf.theta_ = np.array(params['theta'])
clf.sigma_ = np.array(params['sigma'])
clf.classes_ = np.array(params['classes'])
y = clf.predict(X)
return jsonify(pred=y.tolist())
def naiveBayesClassifier(self,X_train,y_train):
"""Method to train a regularized logistic regression classifier
Parameters
----------
X_train: Array shape [n_samples,n_features] for training the model with features
y_train: Array shape[n_samples] for training the model with features of that target
Returns
---------
model: The trained Logistic Regression model."""
meansArray, varianceArray = self.NBMeanAndVariance(X_train,y_train)
model = GaussianNB()
#fit the model with training data
model.fit(X_train,y_train)
model.theta_ = meansArray
model.sigma_ = varianceArray
return model
# In[554]:
#def P9():
### STUDENT END ###
model = GaussianNB()
model.fit(mini_train_data, mini_train_labels)
expected = dev_labels
prediction = model.predict(dev_data)
print "Initial Model"
print classification_report(expected, prediction)
for i in range(1, 100, 5):
model = GaussianNB()
model.fit(mini_train_data, mini_train_labels)
model.sigma_ = model.sigma_ / (i)
expected = dev_labels
prediction = model.predict(dev_data)
acc = accuracy_score(expected, prediction)
plt.scatter(i, acc, color="blue")
plt.title("Accuracy as Sigma Increases")
plt.ylabel("Accuracy"), plt.xlabel("Sigma")
plt.show()
### STUDENT END ###
#gnb = P9()
# ANSWER: ### As shown in the plot above, modifying Sigma can have a dramatic impact on the accuracy outcome.
# (10) Because Naive Bayes is a generative model, we can use the trained model to generate digits. Train a BernoulliNB model and then generate a 10x20 grid with 20 examples of each digit. Because you're using a Bernoulli model, each pixel output will be either 0 or 1. How do the generated digits compare to the training digits?