It's a super deal -- train recurrent network on noisy data and get smooth prediction free
This addresses the problem of noisy time series prediction for neuroscience applications, though it appears incremental as it builds on known properties of recurrent networks.
The paper investigates how predictive recurrent neural networks produce smooth anticipated trajectories from noisy time series input, examining the influence of noise in training data and input sequences on prediction quality. It proposes an explanation for this noise compression phenomenon and discusses its potential significance in neuroscience for understanding evolution in living organisms.
Recent research demonstrate that prediction of time series by predictive recurrent neural networks based on the noisy input generates a smooth anticipated trajectory. We examine influence of the noise component in both the training data sets and the input sequences on network prediction quality. We propose and discuss an explanation of the observed noise compression in the predictive process. We also discuss importance of this property of recurrent networks in the neuroscience context for the evolution of living organisms.