def conditional_mnist(c): N_data, train_images, train_labels, test_images, test_labels = load_mnist() indices = (train_labels[:, c] == 1) c_mean, c_cov = mean_and_cov(train_images[indices, :]) c_model = c_mean, c_cov with open('c_model.pkl', 'w') as f: pickle.dump(c_model, f, 1)
def model_mnist(): # Load and process MNIST data N_data, train_images, train_labels, test_images, test_labels = load_mnist() trained_weights, predict_fun,likeFunc = train_nn(train_images, train_labels, test_images, test_labels) all_mean, all_cov = mean_and_cov(train_images) mnist_models = trained_weights, all_mean, all_cov with open('mnist_models.pkl', 'w') as f: pickle.dump(mnist_models, f, 1)