Online Hyperparameter Meta-Learning with Hypergradient Distillation

Abstract

Many gradient-based meta-learning methods assume a set of parameters that do not participate in inner-optimization, which can be considered as hyperparameters. Although such hyperparameters can be optimized using the existing gradient-based hyperparameter optimization (HO) methods, they suffer from the following issues. Unrolled differentiation methods do not scale well to high-dimensional hyperparameters or horizon length, Implicit Function Theorem (IFT) based methods are restrictive for online optimization, and short horizon approximations suffer from short horizon bias. In this work, we propose a novel HO method that can overcome these limitations, by approximating the second-order term with knowledge distillation. Specifically, we parameterize a single Jacobian-vector product (JVP) for each HO step and minimize the distance from the true second-order term. Our method allows online optimization and also is scalable to the hyperparameter dimension and the horizon length. We demonstrate the effectiveness of our method on two different meta-learning methods and three benchmark datasets.

References in corpus (5)

• Practical Bayesian Optimization of Machine Learning Algorithms
• Theoretical Models of Learning to Learn
• Meta-SGD: Learning to Learn Quickly for Few-Shot Learning
• Cross-Domain Few-Shot Classification via Learned Feature-Wise Transformation
• Online hyperparameter optimization by real-time recurrent learning