skdim.datasets.BenchmarkManifolds

class skdim.datasets.BenchmarkManifolds(random_state: int = None, noise_type: str = 'uniform')

Generates a commonly used benchmark set of synthetic manifolds with known intrinsic dimension described by Hein et al. and Campadelli et al. [Campadelli2015]

Parameters:

noise_type : str, 'uniform' or 'gaussian'

Type of noise to generate

Methods

generate(name, n, dim, d, noise)

Generates all datasets.

generate(name: str = 'all', n: int = 2500, dim: int = None, d: int = None, noise: float = 0.0)

Generates all datasets. A ground truth dict of intrinsic dimension and embedding dimension is in BenchmarkManifolds.dict_truth.keys()

Parameters:

n : int

The number of sample points

dim : int

If generating a single dataset, choose the embedding dimension. Note that some datasets have restrictions on the chosen embedding dimension

d : int

If generating a single dataset, choose the intrinsic dimension. Note that some datasets have restrictions on the chosen intrinsic dimension

noise : float, optional(default=0.0)

The value of noise in data

Returns:

data (a dict of np.arrays or a single np.array with shape (n, dim)) – Generated data