Hi, Is there any documentation or rule of thumb for setting deep learning model hyperparameters based on data frequency? I can't use hyperparameter optimization, so I am looking for some possibly "good" hyperparameter values. I am mainly looking for LSTM, TCN, NBEATS, and NHITS, but it would be nice to have some general references if they exist. Thanks in advance for your time.

The default parameters of all models in neuralforecast are usually good starting points and work well in most cases. You could experiment with different

input_size

. Depending on your horizon, try feeding 2x, 3x or 4x the horizon, and see how it impacts the performance.

Ok thanks I'll try the input size on a subset of the data to see the performance