Deep Learning

LSTM

A recurrent cell with gates that decide what to keep and what to forget.

Long short-term memory adds an explicit cell state plus input, forget and output gates. Information can travel a long way down the sequence without being repeatedly multiplied into nothing.

For a decade this was the default for translation, speech and forecasting. Transformers displaced it for language, but it remains a strong and cheap option for modest time-series problems.

Also in Deep Learning

CNN RNN Transformers Auto Encoders Deep RL Epochs

Apply

Begin the first module

Become AI native, it is the real deal today, and if it is not for you, you have lost nothing but learnt a new skill.

Apply