| import os |
| class Param(): |
| |
| def __init__(self, args): |
| |
| self.hyper_param = self.get_hyper_parameters(args) |
|
|
| def get_hyper_parameters(self, args): |
| """ |
| Args: |
| num_train_epochs_SAE (int): The number of epochs for training stacked auto-encoder. |
| num_train_epochs_DCN (int): The number of epochs for training DCN model. |
| feat_dim (int): The feature dimension. |
| update_interval (int): The number of intervals between contiguous updates. |
| batch_size (int): The batch size for training. |
| lr (float): The learning rate for training DCN. |
| momentum (float): The momentum value of SGD optimizer. |
| tol (float): The tolerance threshold to stop training for DCN. |
| model_name (str): The name of the DCN model (saved in the format of keras). |
| """ |
| hyper_parameters = { |
| 'num_train_epochs_SAE': 5000, |
| 'num_train_epochs_DCN': 12000, |
| 'feat_dim': 2000, |
| 'update_interval': 100, |
| 'DCN_batch_size': 256, |
| 'SAE_batch_size': 4096, |
| 'lr': 0.001, |
| 'momentum': 0.9, |
| 'tol': 0.01, |
| 'model_name': 'DCN.h5' |
| } |
|
|
| return hyper_parameters |
