Train Like a (Var)Pro: Efficient Training of Neural Networks with Variable Projection
This work addresses the problem of slow and inefficient training of neural networks for researchers and practitioners in machine learning and scientific computing, offering an incremental improvement over existing optimization methods.
The paper tackles the challenge of efficiently training deep neural networks by proposing a Gauss-Newton variable projection method (GNvpro) that extends variable projection to non-quadratic loss functions like cross-entropy, enabling faster optimization than stochastic gradient descent in experiments on surrogate modeling, segmentation, and classification tasks.
Deep neural networks (DNNs) have achieved state-of-the-art performance across a variety of traditional machine learning tasks, e.g., speech recognition, image classification, and segmentation. The ability of DNNs to efficiently approximate high-dimensional functions has also motivated their use in scientific applications, e.g., to solve partial differential equations (PDE) and to generate surrogate models. In this paper, we consider the supervised training of DNNs, which arises in many of the above applications. We focus on the central problem of optimizing the weights of the given DNN such that it accurately approximates the relation between observed input and target data. Devising effective solvers for this optimization problem is notoriously challenging due to the large number of weights, non-convexity, data-sparsity, and non-trivial choice of hyperparameters. To solve the optimization problem more efficiently, we propose the use of variable projection (VarPro), a method originally designed for separable nonlinear least-squares problems. Our main contribution is the Gauss-Newton VarPro method (GNvpro) that extends the reach of the VarPro idea to non-quadratic objective functions, most notably, cross-entropy loss functions arising in classification. These extensions make GNvpro applicable to all training problems that involve a DNN whose last layer is an affine mapping, which is common in many state-of-the-art architectures. In our four numerical experiments from surrogate modeling, segmentation, and classification GNvpro solves the optimization problem more efficiently than commonly-used stochastic gradient descent (SGD) schemes. Also, GNvpro finds solutions that generalize well, and in all but one example better than well-tuned SGD methods, to unseen data points.