On the Locality of the Natural Gradient for Deep Learning
This work addresses computational efficiency issues for researchers and practitioners using natural gradient methods in deep learning, though it appears incremental as it builds on existing theory without claiming broad performance gains.
The paper tackles the complexity of applying the natural gradient method in deep Bayesian networks by analyzing two different geometric perspectives and their associated Fisher information matrices, resulting in a proposed method that simplifies computation by incorporating a recognition model.
We study the natural gradient method for learning in deep Bayesian networks, including neural networks. There are two natural geometries associated with such learning systems consisting of visible and hidden units. One geometry is related to the full system, the other one to the visible sub-system. These two geometries imply different natural gradients. In a first step, we demonstrate a great simplification of the natural gradient with respect to the first geometry, due to locality properties of the Fisher information matrix. This simplification does not directly translate to a corresponding simplification with respect to the second geometry. We develop the theory for studying the relation between the two versions of the natural gradient and outline a method for the simplification of the natural gradient with respect to the second geometry based on the first one. This method suggests to incorporate a recognition model as an auxiliary model for the efficient application of the natural gradient method in deep networks.