def test_reviews_bag(self):
"""Classify sentiment using bag-of-words"""
train, dev, test, labels, features = self.split_review_corpus(Review)
classifier = MaxEnt()
classifier.train(train, labels, features, dev)
classifier.save("mikes_model")
class2 = MaxEnt()
class2.load("mikes_model")
self.assertGreater(accuracy(classifier, test), 0.55)
def test_names_nltk(self):
"""Classify names using NLTK features"""
train, dev, test = self.split_names_corpus()
classifier = MaxEnt()
classifier.train(train, dev, 0.1)
acc = accuracy(classifier, test)
self.assertGreater(acc, 0.70)
def test_reviews_bag(self):
"""Classify sentiment using bag-of-words"""
train, dev, test = self.split_review_corpus(BagOfWords)
classifier = MaxEnt()
classifier.train(train, dev)
acc = accuracy(classifier, test)
self.assertGreater(acc, 0.55)
def test_model():
train, dev, test, labels = get_data()
classifier = MaxEnt(train, labels)
classifier.train(train, dev, 0.001, 100)
print("Test result", accuracy(classifier, test))
def test_names_nltk(self):
"""Classify names using NLTK features"""
train, dev, test = self.split_names_corpus()
classifier = MaxEnt()
classifier.labels = ['female', 'male']
classifier.vocab = self.get_names_vocab(train)
classifier.feature_vectors(train + dev + test)
classifier.train(train, dev)
acc = accuracy(classifier, test)
self.assertGreater(acc, 0.70)
def test_reviews_bag(self):
"""Classify sentiment using bag-of-words"""
train, dev, test = self.split_review_corpus(BagOfWords)
classifier = MaxEnt()
classifier.labels = ['positive', 'negative', 'neutral']
classifier.vocab = self.get_review_vocab(train)
print('...creating sparse feature vectors...')
classifier.feature_vectors(train + dev + test)
print('...training...')
classifier.train(train, dev)
self.assertGreater(accuracy(classifier, test), 0.55)
def test_reviews_words(self):
"""Classify sentiment using sentiment words"""
train, dev, test = self.split_review_corpus(SentimentWords)
classifier = MaxEnt()
classifier.train(train, dev)
self.assertGreater(accuracy(classifier, test), 0.55)
self.assertGreater(acc, 0.70)
def split_review_corpus(self, document_class):
"""Split the yelp review corpus into training, dev, and test sets"""
reviews = ReviewCorpus('yelp_reviews.json', document_class=document_class)
seed(hash("reviews"))
shuffle(reviews)
return (reviews[:10000], reviews[10000:11000], reviews[11000:14000])
def test_reviews_bag(self):
"""Classify sentiment using bag-of-words"""
train, dev, test = self.split_review_corpus(BagOfWords)
classifier = MaxEnt()
classifier.train(train, dev)
self.assertGreater(accuracy(classifier, test), 0.55)
'''
if __name__ == '__main__':
docs = Corpus('./gold_standard_all.txt', True)
seed(time.time())
shuffle(docs)
first_80 = round(0.8 * len(docs))
second_10 = first_80 + round(0.1 * len(docs))
train, dev, test = (docs[:first_80], docs[first_80:second_10],
docs[second_10:])
classifier = MaxEnt()
classifier.train(train, dev)
print(accuracy(classifier, test))
def test_reviews_bag(self):
train, dev, test = self.split_review_corpus(BagOfWords)
classifier = MaxEnt()
classifier.train(train, dev)
self.assertGreater(accuracy(classifier, test), 0.55)
def test_names_nltk(self):
train, dev, test = self.split_names_corpus()
classifier = MaxEnt()
classifier.train(train, dev)
acc = accuracy(classifier, test)
self.assertGreater(acc, 0.70)
for i in range(reps):
print(">>>>iteration", i)
reward, states_visited, steps = run_episode(env, valueFunction, n,
False, EPSILON)
#compute feature counts
fcounts = compute_feature_counts(fMap, states_visited, discount, env)
print("steps = ", steps)
#print("feature count = ", fcounts)
features.append(fcounts)
features = np.array(features)
#compute expected feature counts for demos
emp_feature_cnts = np.mean(features[skip_time:], axis=0)
ment = MaxEnt(solve_mdp, fMap, env, num_fcount_rollouts, discount)
maxent_value_fn = ment.get_opt_policy(emp_feature_cnts, learning_rate,
num_steps)
#pickle the controller (value function)
#with open('mcar_maxent_policy_ss.pickle', 'wb') as f:
# pickle.dump(maxent_value_fn, f, pickle.HIGHEST_PROTOCOL)
#with open('mcar_maxent_policy_ss.pickle', 'rb') as f:
# vFunc = pickle.load(f)
#evaluate maxent learned policy
returns = evaluate_softmax_policy(env, eval_rollouts, maxent_value_fn)
print("average return", np.mean(returns))
for r in returns: