Iterated Population Based Training with Task-Agnostic Restarts
Not on the map. Why?
Abstract (arXiv)
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.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 32 of its 49 references that are in the map. First posted on arXiv 12 Nov 2025.
Builds on
32 of 49 references.
References · 32
- 2017Attention Is All You Needcited by 59,932
- 2015Deep Residual Learning for Image Recognitioncited by 51,809
- 2017Proximal Policy Optimization Algorithmscited by 14,441
- 2016SGDR: Stochastic Gradient Descent with Warm Restartscited by 4,993
- 2017Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithmscited by 4,283
- 2022A ConvNet for the 2020scited by 3,545
- 2019RandAugment: Practical automated data augmentation with a reduced search spacecited by 2,261
- 2021EfficientNetV2: Smaller Models and Faster Trainingcited by 1,127
- 2021Deep Reinforcement Learning at the Edge of the Statistical Precipicecited by 781
- 2016Hyperband: A Novel Bandit-Based Approach to Hyperparameter Optimizationcited by 688
- 2018BOHB: Robust and Efficient Hyperparameter Optimization at Scalecited by 466
- 2021Brax -- A Differentiable Physics Engine for Large Scale Rigid Body Simulationcited by 375
20 more
- 2017Population Based Training of Neural Networkscited by 351
- 2021SMAC3: A Versatile Bayesian Optimization Package for Hyperparameter Optimizationcited by 229
- 2015Non-stochastic Best Arm Identification and Hyperparameter Optimizationcited by 219
- 2023The Dormant Neuron Phenomenon in Deep Reinforcement Learningcited by 171
- 2021From Motor Control to Team Play in Simulated Humanoid Footballcited by 87
- 2020Provably Efficient Online Hyperparameter Optimization with Population-Based Banditscited by 61
- 2021Think Global and Act Local: Bayesian Optimisation over High-Dimensional Categorical and Mixed Search Spacescited by 51
- 2016Time-Varying Gaussian Process Bandit Optimizationcited by 50
- 2020Sample-Efficient Automated Deep Reinforcement Learningcited by 31
- 2023PriorBand: Practical Hyperparameter Optimization in the Age of Deep Learningcited by 31
- 2024Weight Clipping for Deep Continual and Reinforcement Learningcited by 29
- 2022Bayesian Generational Population-Based Trainingcited by 20
- 2021Faster Improvement Rate Population Based Trainingcited by 12
- 2022When Does Re-initialization Work?cited by 12
- 2021Tuning Mixed Input Hyperparameters on the Fly for Efficient Population Based AutoRLcited by 8
- 2024Improving Hyperparameter Optimization with Checkpointed Model Weightscited by 5
- 2023Multi-Objective Population Based Trainingcited by 4
- 2025GIT-BO: High-Dimensional Bayesian Optimization with Tabular Foundation Modelscited by 3
- 2023Shrink-Perturb Improves Architecture Mixing during Population Based Training for Neural Architecture Searchcited by 1
- 2025Multiple-Frequencies Population-Based Trainingcited by 1
Built on by
Not cited yet.