Deep Learning the Stock Market
Tal Perry

I’m not sure about your first step — embedding your 4000 dimensional vector into a 300 dimensional space a la word2vec. Word2vec operates under the assumption that “words that appear in the same contexts, are similar”. This is the basis of training the word2vec model. In the case of stocks, its not clear what “the same context means”. Maybe you’ve figured out what “the same context” means for the stocks, but it wasn’t clear from the article. Feel free to ping me if you want to continue to discuss this :)