We studied the effects of time correlation of subsequent patterns on the convergence of on-line learning by a feedforward neural network with the backpropagation algorithm. By using a chaotic time series as sequences of correlated patterns, we found that the unexpected scaling of converging time with the learning parameter emerges when time-correlated patterns accelerate the learning process.
|Number of pages||4|
|Journal||Physical Review E - Statistical Physics, Plasmas, Fluids, and Related Interdisciplinary Topics|
|Publication status||Published - 1996|
ASJC Scopus subject areas
- Statistical and Nonlinear Physics
- Statistics and Probability
- Condensed Matter Physics