Introduction
LSTM stards for Long-short term memory, it is a variant of RNN network. The strong part of LSTM archetect is it has memory build-in. Imaging the daily conversation, we are able to predict next words based on the context of the conversation because we have remembered the important of part of the context. See example below,
ad_pre
What you can remember after several days probabily only a few key information, and you choose to forget irrelavant information. Here is the things you possible remember after several days
ad_post
As LSTM can remember through time. Therefore, it has wide range of applications like text/speech translation, speech to text translation, audio/video prediction
