first implementation of lrDecay, loss and lrCallback, no tests yet

src/helpers.py

@@ -2,7 +2,55 @@ __author__ = 'Lukas Leufen'
 __date__ = '2019-10-21'
 
 
+import logging
+import keras
+import keras.backend as K
+import math
+
+
 def to_list(arg):
     if not isinstance(arg, list):
         arg = [arg]
     return arg
+
+
+class Loss:
+
+    def l_p_loss(self, power):
+        def loss(y_true, y_pred):
+            return K.mean(K.pow(K.abs(y_pred - y_true), power), axis=-1)
+        return loss
+
+
+class lrDecay(keras.callbacks.History):
+
+    def __init__(self, base_lr: float = 0.01, drop: float = 0.96, epochs_drop: int = 8):
+        super(lrDecay, self).__init__()
+
+        self.lr = {'lr': []}
+        self.base_lr = base_lr
+        self.drop = drop
+        self.epochs_drop = epochs_drop
+
+    def on_epoch_begin(self, epoch: int, logs=None):
+        if epoch > 0:
+            current_lr = self.base_lr * math.pow(self.drop, math.floor(1 + epoch) / self.epochs_drop)
+        else:
+            current_lr = self.base_lr
+        K.set_value(self.model.optimizer.lr, current_lr)
+        self.lr['lr'].append(current_lr)
+        logging.info(f"Set learning rate to {current_lr}")
+        return K.get_value(self.model.optimizer.lr)
+
+
+class lrCallback(keras.callbacks.History):
+
+    def __init__(self):
+        super(lrCallback, self).__init__()
+        self.lr = None
+
+    def on_train_begin(self, logs=None):
+        self.lr = {}
+
+    def on_epoch_end(self, epoch, logs=None):
+        self.lr.append(self.model.optimizer.lr)
\ No newline at end of file