hhh920406

class SelfAttention(nn.Module):
    def __init__(self, attention_size, batch_first=False, non_linearity="tanh"):
        super(SelfAttention, self).__init__()

        self.batch_first = batch_first
        self.attention_weights = Parameter(torch.FloatTensor(attention_size))
        self.softmax = nn.Softmax(dim=-1)

        if non_linearity == "relu":
            self.non_linearity = nn.ReLU()

hhh920406

from keras import backend as K, initializers, regularizers, constraints
from keras.engine.topology import Layer


def dot_product(x, kernel):
    """
    Wrapper for dot product operation, in order to be compatible with both
    Theano and Tensorflow
    Args:

hhh920406

- word2vec https://arxiv.org/abs/1310.4546
- sentence2vec, paragraph2vec, doc2vec http://arxiv.org/abs/1405.4053
- tweet2vec http://arxiv.org/abs/1605.03481
- tweet2vec https://arxiv.org/abs/1607.07514
- author2vec http://dl.acm.org/citation.cfm?id=2889382
- item2vec http://arxiv.org/abs/1603.04259
- lda2vec https://arxiv.org/abs/1605.02019
- illustration2vec http://dl.acm.org/citation.cfm?id=2820907
- tag2vec http://ktsaurabh.weebly.com/uploads/3/1/7/8/31783965/distributed_representations_for_content-based_and_personalized_tag_recommendation.pdf
- category2vec http://www.anlp.jp/proceedings/annual_meeting/2015/pdf_dir/C4-3.pdf