SegRNN: Another remarkable work aimed at revitalizing classical RNNs

[lss-1138]

Hi, it's exciting to discover another remarkable work aimed at revitalizing RNN-based methods for time-series analysis tasks.

With the same goal in mind, we have introduced SegRNN (Paper link: https://arxiv.org/abs/2308.11200, Code link: https://github.com/lss-1138/SegRNN), demonstrating that RNNs still excel in time series forecasting tasks.

Specifically, SegRNN leverages two innovative strategies: Segment-wise Iterations and Parallel Multi-step Forecasting, enabling classical RNNs (e.g., GRU) to outperform state-of-the-art Transformer methods with significant efficiency advantages.

Once again, let's work together to make RNNs shine brightly.