All others being equal, would a ResNet-based or RNN-based neural network (with/without an attention mechanism) perform better for forecasting a multivariate time series?
Related:
- Deep learning for time series classification: a review explains that a ResNet performs rather well on multivariate timeseries classification (and I assume thus also on forcasting). Yet a comparrison with a RNN based network is missing.
- Forecasting Multivariate Time Series Data Using Neural Networks (link downloads a .pdf) seems to conclude that a Convolutional Neural Network outperforms a LSTM-based network for time series forecasting.
- CNN vs RNN for time series classification asks the same question about RNN vs CNN (altough for classification, not for forecasting), yet no conclusive answer is given.
- Data Preprocessing and Augmentation for Multiple Short Time Series Forecasting with Recurrent Neural Networks describes a LSTM with skip connections (so basically a RNN/ResNet hybrid) which also seems to perform well.
