mlp-cifar10

Selected MLP trained on CIFAR-10 with clean labels.

Files

• final_mlp_p0.npz: Saved model weights and biases in NumPy NPZ format.
• preprocessing_stats_cifar10_full_train.npz: Training-set preprocessing mean and standard deviation.

Loading the weights

The weights are stored in a custom NumPy NPZ format. They can be inspected using numpy.load; the archive contains weight matrices and bias vectors for the trained network.

The notebook accompanying this repository contains the model implementation and the preprocessing procedure.