kosmos.ml.datasets.mnist_dataset

Classes

class MNISTDataset(*, min_max_scaler: bool = True)

Bases: kosmos.ml.datasets.dataset.SLDataset

MNIST handwritten digits dataset for multiclass classification.

Notes

  • Number of instances: 70,000 (60,000 train + 10,000 test)

  • Number of features: 784 numeric (28x28 pixel images, flattened)

  • Classes: 10 (digits 0-9, roughly balanced)

References

Initialize the dataset.

Parameters:

min_max_scaler (bool) – Whether to apply min-max scaling to the features.


Properties

property class_names

Return human-readable class labels of numbers 0-9.

property input_dimension

Number of feature columns.

property output_dim

Number of distinct classes.