Iterated Population Based Training with Task-Agnostic Restarts

TL;DR

The paper introduces IPBT, a new PBT variant that automatically adjusts hyperparameter restart intervals using task-agnostic restarts and Bayesian optimization, improving performance across various tasks without extra budget.

Contribution

IPBT is a novel PBT method that dynamically tunes restart intervals via task-agnostic restarts and Bayesian optimization, enhancing hyperparameter tuning efficiency.

Findings

IPBT matches or outperforms previous PBT variants.

IPBT outperforms other HPO algorithms like random search, ASHA, SMAC3.

IPBT does not require increased computational budget.

Abstract

Hyperparameter Optimization (HPO) can lift the burden of tuning hyperparameters (HPs) of neural networks. HPO algorithms from the Population Based Training (PBT) family are efficient thanks to dynamically adjusting HPs every few steps of the weight optimization. Recent results indicate that the number of steps between HP updates is an important meta-HP of all PBT variants that can substantially affect their performance. Yet, no method or intuition is available for efficiently setting its value. We introduce Iterated Population Based Training (IPBT), a novel PBT variant that automatically adjusts this HP via restarts that reuse weight information in a task-agnostic way and leverage time-varying Bayesian optimization to reinitialize HPs. Evaluation on 8 image classification and reinforcement learning tasks shows that, on average, our algorithm matches or outperforms 5 previous PBT variants and other HPO algorithms (random search, ASHA, SMAC3), without requiring a budget increase or any changes to its HPs. The source code is available at https://github.com/AwesomeLemon/IPBT.