Perhaps the most powerful tool for combatting the vanishing and exploding gradients in deep neural networks is Batch Normalization. I show how to use tf.keras and TensorFlow to implement Batch Normalization.

View the full post here.