2.5 RNN & LSTM - lvolkmann/CS-490-Deep-Learning GitHub Wiki
Introduction
This week we tackled an old problem with a new approach. We introduced and LSTM layer to a sequential model to aid in context tracking. Additionally, we deployed some other useful tools like GridSearchCV and Tensorboard
Data
These techniques were applied to two data sets:
- sentiments.csv (text, pos/neg/neutral) where text is tweets from the 2016 US election
- spam.csv (text, spam/not spam) where text is random messages
Findings
Conclusion
Ultimately, LSTM is an inexpensive and quick way to improve your net's ability to contextualize data. What is more, GridSearchCV and Tensorboard are some great tools to have in the tool belt.