Optimizing for Generalization in Machine Learning with Cross-Validation Gradients

Abstract

Cross-validation is the workhorse of modern applied statistics and machine learning, as it provides a principled framework for selecting the model that maximizes generalization performance. In this paper, we show that the cross-validation risk is differentiable with respect to the hyperparameters and training data for many common machine learning algorithms, including logistic regression, elastic-net regression, and support vector machines. Leveraging this property of differentiability, we propose a cross-validation gradient method (CVGM) for hyperparameter optimization. Our method enables efficient optimization in high-dimensional hyperparameter spaces of the cross-validation risk, the best surrogate of the true generalization ability of our learning algorithm.

arXiv comment11 pages

References in corpus (5)

• Scikit-learn: Machine Learning in Python
• Practical Bayesian Optimization of Machine Learning Algorithms
• Gradient-based Hyperparameter Optimization through Reversible Learning
• OptNet: Differentiable Optimization as a Layer in Neural Networks
• On the Differentiability of the Solution to Convex Optimization Problems