def __init__(self,
batch_size,
embedding_layer: layers.Embedding = None,
activation=None,
**kwargs):
super().__init__(**kwargs)
self.embedding_layer = embedding_layer
self.vocab_size = embedding_layer.get_config()['input_dim']
self.activation = activations.get(activation)
self.trainable = False
self.batch_size = batch_size
# In[15]:
tokens = []
tokenizer = RegexpTokenizer(r'\w+')
for i, s in enumerate(speeches['speech']):
tokens.append(tokenizer.tokenize(s))
if i % 10000 == 0:
# print(i)
# In[17]:
n_fact = embedding_layer.get_config()['output_dim']
idx_tokens = [[vocab[word]
for word in doc if word in vocab.keys()] \
for doc in tokens]
# In[21]:
from sklearn.cross_validation import train_test_split
train_idxs, test_idxs, y_train, y_test = train_test_split(speeches.index,
np.where(speeches['party'] == 'R', 1, 0),
test_size=0.2)
# In[22]:
