Random Projection Recurrent Neural Networks for Time Series Classification
Ye Ma, Moufa Hu, Qing Chang, Huanzhang Lu · 2018
Recurrent neural networks (RNNs) remain challenging and keep lack of a long-term memory or learning ability in sequential data classification and prediction. In this paper, we propose an ingenious recurrent model, Random projection RNNs (R-RNNs), incorporating random projection to advance the classification accuracy limited by the gradient vanishing and exploding problem before. Experiments on public time series datasets demonstrate that our proposed method outperforms several existing typical models substantially. And the efficiency of R-RNNs concerning SNR is also discussed.