ctf4science.data_module.load_validation_dataset#
- ctf4science.data_module.load_validation_dataset(dataset_name, pair_id, train_split=0.8, transpose=False)#
Load dataset and split into training, validation, and initialization.
For pair_ids 2 and 4, validation is a copy of training. For pair_ids 8 and 9, one training matrix becomes validation and init_data is the burn-in subset; for 1, 3, 5, 6, 7, the single matrix is split by train_split.
- Parameters:
- dataset_namestr
Name of the dataset to load.
- pair_idint
Data pair (1–9; special handling for 2, 4, 8, 9).
- train_splitfloat, optional
Fraction of data for training in (0, 1). Default is 0.8. Ignored for pair_ids 2, 4, 8, 9.
- transposebool, optional
If True, return arrays as (features, timesteps); otherwise (timesteps, features). Default is False.
- Returns:
- train_datalist of ndarray
Training data arrays (after split).
- val_datandarray
Validation data matrix.
- init_datandarray or None
Initialization (burn-in) data; only non-None for pair_id 8 and 9.
- Raises:
- ValueError
If train_split is not in (0, 1), pair_id is not in 1–9, or pair_id 8/9 has no initialization data.