Explore efficient neural network architectures for sequence processing. Learn about innovative methods that reduce computational demands and improve performance in handling long time lags through self-organizing multilevel hierarchies and collapsing predictor hierarchies.