Robust Implicit Backpropagation
This addresses the problem of unstable training due to learning rate sensitivity for neural network practitioners, representing a novel application rather than an incremental improvement.
The paper tackles the challenge of neural network hyperparameter tuning, particularly sensitivity to learning rate, by applying Implicit Stochastic Gradient Descent (ISGD) for the first time in this context, resulting in a method that is more robust to high learning rates and generally outperforms standard backpropagation across various tasks.
Arguably the biggest challenge in applying neural networks is tuning the hyperparameters, in particular the learning rate. The sensitivity to the learning rate is due to the reliance on backpropagation to train the network. In this paper we present the first application of Implicit Stochastic Gradient Descent (ISGD) to train neural networks, a method known in convex optimization to be unconditionally stable and robust to the learning rate. Our key contribution is a novel layer-wise approximation of ISGD which makes its updates tractable for neural networks. Experiments show that our method is more robust to high learning rates and generally outperforms standard backpropagation on a variety of tasks.