Pushing Stochastic Gradient towards Second-Order Methods -- Backpropagation Learning with Transformations in Nonlinearities
This work addresses the challenge of inefficient training in neural networks for machine learning practitioners, though it is incremental as it builds on prior transformations.
The authors tackled the problem of slow learning in multi-layer perceptrons by introducing transformations to normalize hidden neuron outputs, showing that this makes stochastic gradient descent behave more like second-order optimization methods and speeds up learning, with experiments indicating a trade-off between faster convergence and potential performance degradation.
Recently, we proposed to transform the outputs of each hidden neuron in a multi-layer perceptron network to have zero output and zero slope on average, and use separate shortcut connections to model the linear dependencies instead. We continue the work by firstly introducing a third transformation to normalize the scale of the outputs of each hidden neuron, and secondly by analyzing the connections to second order optimization methods. We show that the transformations make a simple stochastic gradient behave closer to second-order optimization methods and thus speed up learning. This is shown both in theory and with experiments. The experiments on the third transformation show that while it further increases the speed of learning, it can also hurt performance by converging to a worse local optimum, where both the inputs and outputs of many hidden neurons are close to zero.