A recurrent neural network for modelling dynamical systems

We introduce a recurrent network architecture for modelling a general class of dynamical systems. The network is intended for modelling real-world pro cesses in which empirical measurements of the external and state variables are obtained-at discrete time points. The model can learn from multiple tem poral patterns, which may evolve on different timescales and be sampled at non-uniform time intervals. We demonstrate the application of the model to a synthetic problem in which target data are only provided at the final tim e step. Despite the sparseness of the training data, the network is able no t only to make good predictions at the final time step for temporal process es unseen in training, but also to reproduce the sequence of the state vari ables at earlier times. Moreover, we show how the network can infer the exi stence and role of state variables for which no target information is provi ded. The ability of the model to cope with sparse data is likely to be usef ul in a number of applications, including, in particular, the modelling of metal forging.