Apache Spark Deep Learning Cookbook

How to do it...

This section will describe the exploding gradient problem in recurrent neural networks:

• The exploding gradient problem is very similar to the vanishing gradient problem, but just the opposite.
• When long-term dependencies arise in recurrent neural networks, the error term is propagated backward through the network sometimes explodes or becomes very large.
• This error term multiplied by the learning rate results in an extremely large w. This gives rise to new weights that look very different from the previous weights. It is called the exploding gradient problem because the value of the gradient becomes too large.
• The problem of ...