For the sake of comparison, we should train simple models for hysteresis prediction to compare against the more complex TFT. This is useful for academic reasons, but also to justify the choice of the model architecture. We should also include examples with or without the time axis.
- Train a basic MLP for hysteresis prediction [priority:: low] [completion:: 2025-08-01]
- Train a TSMixer for hysteresis prediction [ priority:: low] [completion:: 2025-08-01]
- Train a basic LSTM for hysteresis prediction [priority:: low] [completion:: 2025-08-01]
Evaluation metrics
Compare using RMSE and SMAPE.