Return delta of FedMD-NFDP Args: n (int): training set size k (int): sampling size replacement (bool, optional): sampling

Get noise multiplier. Args: noise_multiplier_key (str): Key for noise multiplier. target_epsilon (float): Target epsi

Get similar boxes between two frames based on the peak signal-to-noise ratio (PSNR). Args: frame_1 (numpy.ndarray): The first frame.

* @brief Gets the span of partitioned dataframe * * @param df * @param partition * @param scale_map * @return std::map<std::string, float> */

* @brief Inserts a continuous value * * @param column * @param value */

FOR INTERNAL USE ONLY

* @brief Checks whether the dataframe and partition meet k-anonymous property * * @param df * @param partition * @param sensitive_column * @param

Compute loss. Args: y (torch.Tensor): Target tensor. result_dict (dict): Dictionary containing sampled outpu

* @brief Return the predicted value of the give new sample X * * @param X the new sample to be predicted * @return vector<vector<float>> *

get probabilities of whether the given prediction came from training data or not Args: y_pred_prob (torch.Tensor): pre

Privatizes the optimizer. Args: noise_multiplier (float): Noise multiplier for privacy. Returns: tu

Receive the local models Args: use_gradients (bool, optional): If True, receive the local gradients. Otherwise, receive the local

Sample noise for the given input. Args: x (torch.Tensor): Input tensor. Returns: torch.Tensor: Samp

Update the parameters with the give gradient Args: grads (List[torch.Tensor]): list of gradients

Update the parameters with the give gradient Args: grads (List[torch.Tensor]): list of gradients

Update the global model Args: use_gradients (bool, optional): If True, update the global model with aggregated local gradients. D

Upload the output logits on the public dataset to the server. Returns: torch.Tensor: the output logits of the public dataset.

Upload the model parameters

Initialize a FlickerException. Args: message (str): The error message.

Initialize the BaseMomentAccountant. Args: search (str, optional): The search strategy. Defaults to "ternary".

Apply side information for Adam optimizer. Args: opt: Optimizer instance.

Apply side information for RMSprop optimizer. Args: opt: Optimizer instance.

calculates the regularization term Args: fake_x: reconstructed images fake_pred: the predicted value of reconstructed

Client-side back propagation Args: grad_from_server: gradient which the server send to the client

Sets proper attributes based on the type of classifier Returns: return True (bool) if there is no error Raises:

detect the type of classifier and prepare proper settings Returns: return true if no error occurs Raises: Va

Aggregate the received logits with ERA

Estimates the secret labels from the received gradients this function is based on the following papers: batch_size == 1: https://arxiv.org/ab

Trains shadow models on given data Args: X (np.array): training data for shadow models y (np.array): training label f

Calculates deviation of the estimated density p(xm−1 |yc = −1) Args: xm (np.array) : an adversarial example Returns:

Calculates deviation of objective function F Args: xm (np.array) : an adversarial example norm (str) : type of distan

Returns a hook function to extract the input of the specified layer of the target model Args: name: the key of self.bn_reg_layer_

Inits the fake labels Args: batch_size: the batch size Returns: randomly initialized or estimated labels

Inits the fake images Args: batch_size: the batch size Returns: randomly generated torch.Tensor whose shape is (batch_size,