Learning the Regularization Strength for Deep Fine-Tuning via a Data-Emphasized Variational Objective
This addresses the computational inefficiency and data waste in hyperparameter tuning for practitioners using deep fine-tuning, though it is incremental as it builds on existing variational methods.
The paper tackles the problem of expensive grid search for regularization hyperparameters in transfer learning by proposing a method to directly learn these hyperparameters using a modified evidence lower bound (ELBo) objective, achieving comparable heldout accuracy on image classification tasks with significantly reduced compute time.
A number of popular transfer learning methods rely on grid search to select regularization hyperparameters that control over-fitting. This grid search requirement has several key disadvantages: the search is computationally expensive, requires carving out a validation set that reduces the size of available data for model training, and requires practitioners to specify candidate values. In this paper, we propose an alternative to grid search: directly learning regularization hyperparameters on the full training set via model selection techniques based on the evidence lower bound ("ELBo") objective from variational methods. For deep neural networks with millions of parameters, we specifically recommend a modified ELBo that upweights the influence of the data likelihood relative to the prior while remaining a valid bound on the evidence for Bayesian model selection. Our proposed technique overcomes all three disadvantages of grid search. We demonstrate effectiveness on image classification tasks on several datasets, yielding heldout accuracy comparable to existing approaches with far less compute time.