New publication
"Modified Deep Transformers for GNSS Time Series Prediction" by Kiani Shahvandi and Soja (2021)
In this contribution, Mostafa Kiani Shahvandi and Benedikt Soja investigated the problem of GNSS station coordinate time series prediction. They modified the state-of-the-art Transformer model to make it suitable for the sequence-to-sequence regression tasks. By applying the revised architecture to a large number of GNSS stations across the globe, they demonstrated that it can outperform other methods (including recurrent and non-recurrent neural networks) in terms of prediction performance.
Have a look at this exciting publication: external page https://ieeexplore.ieee.org/document/9554764