Closed joetigger closed 8 years ago
All current RNN models are stateless, i.e. when calling net.predict() it starts all over with clean hidden state. Is there any way that I can call RNN without resetting hidden state? This is very useful when predicting next value in a time series.
This is not currently possible with theanets, but there is an open issue to keep track of it -- see #81.
All current RNN models are stateless, i.e. when calling net.predict() it starts all over with clean hidden state. Is there any way that I can call RNN without resetting hidden state? This is very useful when predicting next value in a time series.