def main():
text = 'You say goodbye and I say hello.'
corpus, word_to_id, id_to_word = preprocess(text)
vocab_size = len(word_to_id)
C = create_co_matrix(corpus, vocab_size)
W = ppmi(C)
np.set_printoptions(precision=3) # 有効桁3桁で表示
print('convariance matrix')
print(C)
print('-'*50)
print('PPMI')
print(W)
def test_ppmi(self):
text = 'you say goodbye and I say hello.'
corpus, w2id, id2w = preprocess(text)
vocab_size = len(w2id)
C = create_co_matrix(corpus, vocab_size)
W = ppmi(C)
W = np.around(W, 3)
expected = np.array([[0., 1.807, 0., 0., 0., 0., 0.],
[1.807, 0., 0.807, 0., 0.807, 0.807, 0.],
[0., 0.807, 0., 1.807, 0., 0., 0.],
[0., 0., 1.807, 0., 1.807, 0., 0.],
[0., 0.807, 0., 1.807, 0., 0., 0.],
[0., 0.807, 0., 0., 0., 0., 2.807],
[0., 0., 0., 0., 0., 2.807, 0.]])
np.testing.assert_array_almost_equal(W, expected)
def main():
text = 'You say goodbye and I say hello.'
corpus, word_to_id, id_to_word = preprocess(text)
vocab_size = len(word_to_id)
C = create_co_matrix(corpus, vocab_size, window_size=1)
W = ppmi(C)
# SVD
U, S, V = np.linalg.svd(W)
print(C[0]) # 共起行列
print(W[0]) # PPMI行列
print(U[0]) # SVD
print(U[0, :2])
for word, word_id in word_to_id.items():
plt.annotate(word, (U[word_id, 0], U[word_id, 1]))
plt.scatter(U[:, 0], U[:, 1], alpha=0.5)
plt.show()
import sys
sys.path.append('..')
import numpy as np
from common.util import preprocess, create_co_matrix, cos_similarity, ppmi
text = "You say goodbye and I say hello."
corpus, word_to_id, id_to_word = preprocess(text)
vocab_size = len(word_to_id)
C = create_co_matrix(corpus, vocab_size)
W = ppmi(C)
np.set_printoptions(precision=3) # 유효 자릿수를 세 자리로 표시
print('동시발생 행렬')
print(C)
print('-' * 50)
print('PPMI')
print(W)
# coding: utf-8
import sys
sys.path.append('/home/hiromasa/deep-learning-from-scratch-2')
import numpy as np
from common.util import most_similar, create_co_matrix, ppmi
from dataset import ptb
window_size = 2
wordvec_size = 100
corpus, word_to_id, id_to_word = ptb.load_data('train')
vocab_size = len(word_to_id)
print('counting co-occurrence /home/hiromasa/deep-learning-from-scratch-2.')
C = create_co_matrix(corpus, vocab_size, window_size)
print('calculating PPMI /home/hiromasa/deep-learning-from-scratch-2.')
W = ppmi(C, verbose=True)
print('calculating SVD /home/hiromasa/deep-learning-from-scratch-2.')
try:
# truncated SVD (fast!)
from sklearn.utils.extmath import randomized_svd
U, S, V = randomized_svd(W,
n_components=wordvec_size,
n_iter=5,
random_state=None)
except ImportError:
# SVD (slow)
U, S, V = np.linalg.svd(W)
word_vecs = U[:, :wordvec_size]
for j in range(C.shape[1]):
pmi = np.log2(C[i, j] * N / (S[j] * S[i]) + eps)
M[i, j] = max(0, pmi)
if verbose:
cnt += 1
if cnt % (total // 100) == 0:
print('%.1f%% done' % (100 * cnt / total))
return M
text = 'You say goodbye and I say hello.'
corpus, word_to_id, id_to_word = preprocess(text)
vocab_size = len(word_to_id)
C = create_co_matrix(corpus, vocab_size)
W = ppmi(C)
np.set_printoptions(precision=3) # 有効桁3桁で表示
print('covariance matrix')
print(C)
# [[0 1 0 0 0 0 0]
# [1 0 1 0 1 1 0]
# [0 1 0 1 0 0 0]
# [0 0 1 0 1 0 0]
# [0 1 0 1 0 0 0]
# [0 1 0 0 0 0 1]
# [0 0 0 0 0 1 0]]
print('-' * 50)
print('PPMI')
print(W)
# [[0. 1.807 0. 0. 0. 0. 0. ]