相关疑难解决方法(0)

我有一些我希望通过Doc2Vec模型运行的示例句子.我的最终目标是一个大小矩阵(num_sentences,num_features).

我正在使用Gensim包.

from gensim.models.doc2vec import TaggedDocument
from gensim.models import Doc2Vec
# warning: long sample of data. It's just 40 sentences really though.
labeled_sents = [TaggedDocument(words=['u0644', 'u0646', 'u062f', 'u0646', 'u060c', 'u0628', 'u0631', 'u0637', 'u0627', 'u0646', 'u06cc', 'u06c1', 'u06a9', 'u0627'], tags='400'), TaggedDocument(words=['do', 'pan', 'en', '1713', 'o', 'soar', 'onde', 'se', 'sit', 'xfaa'], tags='401'), TaggedDocument(words=['u0420', 'u044c', 'u043e', 'u043d', 'u0442', 'u0433', 'u0435', 'u043d', '1901', 'xa0', 'u2022', 'u041b', 'u043e', 'u0440', 'u0435', 'u043d', 'u0446', 'xa0', 'u0417', 'u0435', 'u0435', 'u043c', 'u0430', 'u043d', '1902', 'xa0', 'u2022', 'u0411', 'u0435', 'u043a', …

我试图理解函数中的epochs参数Doc2Vec和epochs函数中的参数train。

在以下代码片段中，我手动设置了4000次迭代的循环。是否需要或在Doc2Vec中传递4000作为纪元参数足够？另外epochsin Doc2Vec与时代in有train什么不同？

documents = Documents(train_set)

model = Doc2Vec(vector_size=100, dbow_words=1, dm=0, epochs=4000,  window=5,
                seed=1337, min_count=5, workers=4, alpha=0.001, min_alpha=0.025)

model.build_vocab(documents)

for epoch in range(model.epochs):
    print("epoch "+str(epoch))
    model.train(documents, total_examples=total_length, epochs=1)
    ckpnt = model_name+"_epoch_"+str(epoch)
    model.save(ckpnt)
    print("Saving {}".format(ckpnt))

此外，权重如何以及何时更新？