我发现了Kim关于CNN使用word2vec的演示文稿幻灯片:http://www.people.fas.harvard.edu/~yoonkim/data/Kim_EMNLP_2014_slides.pdf
在第20张幻灯片上,第四个要点如下:
Words not in word2vec are initialized randomly from U[−a, a]
where a is chosen such that the unknown words have the
same variance as words already in word2vec.
现在我想知道如何计算“a”,以及如何计算完全未知单词的整个向量。