Autoencoders for complex- and real-valued autoencoder

This repositorio contains the code related to paper "GNSS anomaly detection with complex-valued LSTM-based autoencoder":

O. Savolainen, A. Elango, A. Morrison, N. Sokolova and L. Ruotsalainen, "GNSS Anomaly Detection with Complex-Valued LSTM Networks," 2024 International Conference on Localization and GNSS (ICL-GNSS), Antwerp, Belgium, 2024, pp. 1-7, doi: 10.1109/ICL-GNSS60721.2024.10578405.

Data

The data reading function expects the binary file(s) (.DAT or .dat) in the I/Q format. The model should be trained only with clean data.

Running the code

1. Choose either run_complex.py or run_real.py
2. Adjust hyperparameters and file names
3. Run the code