Time-RNNs

Pytorch Implementation of Time-LSTM as shown in "What to Do Next: Modeling User Behaviors by Time-LSTM" and implementation of Time-GRU as shown in "Is Timing Critical to Trace Reconstruction?". The addition of the Time Gate can be generalized to other Recurrent Neural Networks as well. In terms of performance, all RNNs with a cuantitative notion of time (Time RNN) achieve better results than RNNs with a qualitative notion of time (standard RNNs). We experiment with traces of the Blackberry QNX system operating in a car.