Probabilistic Rollouts for Learning Curve Extrapolation Across Hyperparameter Settings

Code

• github.com/gmatilde/vdrnn
  OfficialIn paperpytorch★ 0

Abstract

We propose probabilistic models that can extrapolate learning curves of iterative machine learning algorithms, such as stochastic gradient descent for training deep networks, based on training data with variable-length learning curves. We study instantiations of this framework based on random forests and Bayesian recurrent neural networks. Our experiments show that these models yield better predictions than state-of-the-art models from the hyperparameter optimization literature when extrapolating the performance of neural networks trained with different hyperparameter settings.

Tasks

Reproductions