Noisy Recurrent Neural Networks

The dataset is a class of noisy recurrent neural networks with w (unbounded) weights for classification of sequences of length T, where independent noise distributed according to N (0, σ2) is added to the output of each neuron in the network.