Temporal pattern learning in noisy recurrent neural networks
Suddhasattwa Das, O. Olurotimi · 2002
One of the important applications of recurrent neural networks (RNN) is in generating temporal patterns. This is relevant in many dynamic system identification and modeling problems. Since noisy input is common, a quantitative analysis of temporal pattern generation in the presence of noise is essential. In our previous work we established a number of quantitative measures of noisy RNN performance. This paper demonstrates their application to a trajectory generation problem.