[doc] Add referencing to each regularization techniques

Author: nabenabe0928

autoPyTorch/pipeline/components/training/trainer/AdversarialTrainer.py

@@ -21,6 +21,13 @@
 
 
 class AdversarialTrainer(BaseTrainerComponent):
+    """
+    References:
+        Title: Explaining and Harnessing Adversarial Examples
+        Authors: Ian J. Goodfellow et. al.
+        URL: https://arxiv.org/pdf/1412.6572.pdf
+        Github URL: https://pytorch.org/tutorials/beginner/fgsm_tutorial.html#fgsm-attack
+    """
     def __init__(
             self,
             epsilon: float,
@@ -37,11 +44,7 @@ def __init__(
 
         Args:
             epsilon (float): The perturbation magnitude.
-        
-        References:
-            Explaining and Harnessing Adversarial Examples
-            Ian J. Goodfellow et. al.
-            https://arxiv.org/pdf/1412.6572.pdf
+
         """
         super().__init__(random_state=random_state,
                          weighted_loss=weighted_loss,
@@ -129,9 +132,6 @@ def fgsm_attack(
 
         Returns:
             adv_data (np.ndarray): the adversarial examples.
-        
-        References:
-            https://pytorch.org/tutorials/beginner/fgsm_tutorial.html#fgsm-attack
         """
         data_copy = deepcopy(data)
         data_copy = data_copy.float().to(self.device)

autoPyTorch/pipeline/components/training/trainer/GridCutMixTrainer.py

@@ -9,6 +9,14 @@
 
 
 class GridCutMixTrainer(MixUp, BaseTrainerComponent):
+    """  # noqa
+    References:
+        Title: CutMix: Regularization Strategy to Train Strong Classifiers
+               with Localizable Features
+        Authors: Sangdoo Yun et. al.
+        URL: https://openaccess.thecvf.com/content_ICCV_2019/papers/Yun_CutMix_Regularization_Strategy_to_Train_Strong_Classifiers_With_Localizable_Features_ICCV_2019_paper.pdf
+        Github URL: https://github.com/clovaai/CutMix-PyTorch/blob/master/train.py#L227-L244
+    """
 
     def data_preparation(self, X: np.ndarray, y: np.ndarray,
                          ) -> typing.Tuple[np.ndarray, typing.Dict[str, np.ndarray]]:

autoPyTorch/pipeline/components/training/trainer/GridCutOutTrainer.py

@@ -7,6 +7,13 @@
 
 
 class GridCutOutTrainer(CutOut, BaseTrainerComponent):
+    """
+    References:
+        Title: Improved Regularization of Convolutional Neural Networks with Cutout
+        Authors: Terrance DeVries and Graham W. Taylor
+        URL: https://arxiv.org/pdf/1708.04552.pdf
+        Github URL: https://github.com/hysts/pytorch_cutout/blob/master/dataloader.py#L36-L68
+    """
 
     def data_preparation(self, X: np.ndarray, y: np.ndarray,
                          ) -> typing.Tuple[np.ndarray, typing.Dict[str, np.ndarray]]:

autoPyTorch/pipeline/components/training/trainer/RowCutOutTrainer.py

@@ -9,6 +9,14 @@
 
 
 class RowCutOutTrainer(CutOut, BaseTrainerComponent):
+    """
+    References:
+        Title: Improved Regularization of Convolutional Neural Networks with Cutout
+        Authors: Terrance DeVries and Graham W. Taylor
+        URL: https://arxiv.org/pdf/1708.04552.pdf
+        Github URL: https://github.com/hysts/pytorch_cutout/blob/master/dataloader.py#L36-L68
+    """
+
     # 0 is non-informative in image data
     NUMERICAL_VALUE = 0
     # -1 is the conceptually equivalent to 0 in a image, i.e. 0-pad

autoPyTorch/pipeline/components/training/trainer/mixup_utils.py

@@ -18,6 +18,13 @@
 
 
 class MixUp:
+    """
+    References:
+        Title: mixup: Beyond Empirical Risk Minimization
+        Authors: Hougyi Zhang et. al.
+        URL: https://arxiv.org/pdf/1710.09412.pdf%C2%A0
+        Github URL: https://github.com/facebookresearch/mixup-cifar10/blob/master/train.py#L119-L138
+    """
     def __init__(self, alpha: float,
                  weighted_loss: bool = False,
                  random_state: Optional[np.random.RandomState] = None,