Amortized Safe Active Learning for Real-Time Data Acquisition: Pretrained Neural Policies From Simulated Nonparametric Functions
Not on the map. Why?
Abstract (arXiv)
Safe active learning (AL) is a sequential scheme for learning unknown systems while respecting safety constraints during data acquisition. Existing methods often rely on Gaussian processes (GPs) to model the task and safety constraints, requiring repeated GP updates and constrained acquisition optimization--incurring significant computations which are challenging for real-time decision-making. We propose amortized AL for regression and amortized safe AL, replacing expensive online computations with a pretrained neural policy. Inspired by recent advances in amortized Bayesian experimental design, we leverage GPs as pretraining simulators. We train our policy prior to the AL deployment on simulated nonparametric functions, using Fourier feature-based GP sampling and a differentiable acquisition objective that is safety-aware in the safe AL setting. At deployment, our policy selects informative and (if desired) safe queries via a single forward pass, eliminating GP inference and acquisition optimization. This leads to magnitudes of speed improvements while preserving learning quality. Our framework is modular and, without the safety component, yields fast unconstrained AL for time-sensitive tasks.
Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Builds on 23 of its 55 references that are in the map. First posted on arXiv 26 Jan 2025.
Builds on
23 of 55 references.
References · 23
- 2017Attention Is All You Needcited by 60,361
- 2019On the Variance of the Adaptive Learning Rate and Beyondcited by 772
- 2018Pyro: Deep Universal Probabilistic Programmingcited by 380
- 2021Safe Learning in Robotics: From Learning-Based Control to Safe Reinforcement Learningcited by 371
- 2014Scalable Variational Gaussian Process Classificationcited by 228
- 2016Bayesian Optimization with Safety Constraints: Safe and Automatic Parameter Tuning in Roboticscited by 134
- 2020Efficiently Sampling Functions from Gaussian Process Posteriorscited by 127
- 2023Modern Bayesian Experimental Designcited by 125
- 2014Bayesian Optimization with Unknown Constraintscited by 109
- 2021Deep Adaptive Design: Amortizing Sequential Bayesian Experimental Designcited by 106
- 2020PACOH: Bayes-Optimal Meta-Learning with PAC-Guaranteescited by 85
- 2018Stagewise Safe Bayesian Optimization with Gaussian Processescited by 79
11 more
- 2015Safe Controller Optimization for Quadrotors with Gaussian Processescited by 78
- 2019Adaptive and Safe Bayesian Optimization in High Dimensions via One-Dimensional Subspacescited by 73
- 2021Implicit Deep Adaptive Design: Policy-Based Experimental Design without Likelihoodscited by 61
- 2023Inducing Point Allocation for Sparse Gaussian Processes in High-Throughput Bayesian Optimisationcited by 32
- 2024Amortized Bayesian Experimental Design for Decision-Makingcited by 21
- 2022Safe Active Learning for Multi-Output Gaussian Processescited by 13
- 2024Safe Active Learning for Time-Series Modeling with Gaussian Processescited by 12
- 2015Near-Optimal Active Learning of Multi-Output Gaussian Processescited by 10
- 2023Amortized Inference for Gaussian Process Hyperparameters of Structured Kernelscited by 10
- 2022Information-Theoretic Safe Exploration with Gaussian Processescited by 9
- 2024Global Safe Sequential Learning via Efficient Knowledge Transfercited by 1
Built on by
Citing papers per year: 2026: 3.
3 papers.