kaira.data.datasets.CorrelatedDataset

Inheritance diagram for CorrelatedDataset

class kaira.data.datasets.CorrelatedDataset(length: int, shape: int | Tuple[int, ...] = (128,), correlation: float = 0.8, noise_std: float = 0.1, seed: int | None = None)

Bases: Dataset

Dataset for correlated data pairs.

Generates pairs of correlated tensors useful for Wyner-Ziv coding, side information experiments, and correlation modeling.

Methods

__init__

Initialize the correlated dataset.

__init__(length: int, shape: int | Tuple[int, ...] = (128,), correlation: float = 0.8, noise_std: float = 0.1, seed: int | None = None)

Initialize the correlated dataset.

Parameters:

• length – Number of samples in the dataset

• shape – Shape of each tensor (int for 1D, tuple for multi-dimensional)

• correlation – Correlation coefficient between source and side info (0-1)

• noise_std – Standard deviation of noise added to create correlation

• seed – Random seed for reproducibility