ctf4science.data_module.get_prediction_timesteps

ctf4science.data_module.get_prediction_timesteps(dataset_name: str, pair_id: int, subset: str = 'test') → ndarray

Return physical time values at which predictions must be evaluated.

Computes absolute physical times using the formula:

timesteps[i] = (start_index + i) * delta_t,  i = 0, ..., N-1

where start_index = metadata['matrix_start_index'][matrix_name], N = metadata['matrix_shapes'][matrix_name][0] (number of rows), and delta_t = metadata['delta_t']. These are absolute times along the underlying trajectory, not zero-based indices. The first value is start_index * delta_t and the last is (start_index + N - 1) * delta_t.

Parameters:

dataset_namestr

Name of the dataset (e.g. 'ODE_Lorenz', 'PDE_KS').

pair_idint

ID of the train-test pair.

subset{‘test’, ‘initialization’}, optional

Which matrix to use. Default is 'test'.

Returns:

ndarray

1D array of length N with physical time values spanning [start_index * delta_t, (start_index + N - 1) * delta_t].

Raises:

ValueError

If pair_id is missing, subset matrix or metadata (e.g. matrix_shapes, matrix_start_index, delta_t) is missing.

Notes

Example — ODE_Lorenz, pair 1, subset=’test’: X1test.mat has shape [1000, 3], start_index=10000, delta_t=0.05. Returns 1000 values in [500.0, 549.95].

Example — ODE_Lorenz, pair 6, subset=’test’: X6test.mat has shape [1000, 3], start_index=100, delta_t=0.05. Returns 1000 values in [5.0, 54.95].

See the dataset YAML configs under data/<dataset_name>/ for the matrix_shapes, matrix_start_index, and delta_t values for every matrix.