decent_bench.datasets

class decent_bench.datasets.A

Feature matrix type.

alias of ndarray[tuple[Any, …], dtype[float64]]

class decent_bench.datasets.b

Target vector type.

alias of ndarray[tuple[Any, …], dtype[float64]]

class decent_bench.datasets.DatasetPartition

Tuple of (A, b) representing one dataset partition.

alias of tuple[A, b]

class decent_bench.datasets.Dataset

Bases: ABC

Dataset containing partitions in the form of feature matrix A and target vector b.

abstractmethod get_training_partitions() → Sequence[DatasetPartition]

Partitions used for finding the optimal optimization variable x.

class decent_bench.datasets.SyntheticClassificationData(n_partitions: int, n_classes: int, n_samples_per_partition: int, n_features: int, seed: int | None = None)

Bases: Dataset

Dataset with synthetic classification data.

Parameters:

• n_partitions – number of training partitions to generate, i.e. the length of the sequence returned by get_training_partitions()

• n_classes – number of classes, i.e. unique values in target vector b

• n_samples_per_partition – number of rows in A and b per partition

• n_features – columns in A

• seed – used for random generation, set to a specific value for reproducible results

get_training_partitions() → list[DatasetPartition]

Partitions used for finding the optimal optimization variable x.