Unsupervised Continual Learning for Amortized Bayesian Inference

2026-02-26

AI Summary

Amortized Bayesian inference suffers significant performance degradation under model misspecification or distributional shift, and existing self-consistent training approaches struggle to adapt to continuously arriving data. This work introduces unsupervised continual learning into this framework for the first time, proposing a decoupled continual learning strategy that combines simulation-based pretraining with sequential self-consistent fine-tuning, experience replay, and Elastic Weight Consolidation (EWC) to mitigate catastrophic forgetting. Evaluated on three benchmark cases, the proposed method effectively alleviates forgetting and yields posterior estimates that substantially outperform those from standard simulation-based training, achieving closer alignment with reference Markov Chain Monte Carlo (MCMC) results.

Technology Category

• Machine Learning: Life-Long and Continual Learning
• Reasoning under Uncertainty: Sequential Decision Making
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Amortized Bayesian Inference (ABI) enables efficient posterior estimation using generative neural networks trained on simulated data, but often suffers from performance degradation under model misspecification. While self-consistency (SC) training on unlabeled empirical data can enhance network robustness, current approaches are limited to static, single-task settings and fail to handle sequentially arriving data or distribution shifts. We propose a continual learning framework for ABI that decouples simulation-based pre-training from unsupervised sequential SC fine-tuning on real-world data. To address the challenge of catastrophic forgetting, we introduce two adaptation strategies: (1) SC with episodic replay, utilizing a memory buffer of past observations, and (2) SC with elastic weight consolidation, which regularizes updates to preserve task-critical parameters. Across three diverse case studies, our methods significantly mitigate forgetting and yield posterior estimates that outperform standard simulation-based training, achieving estimates closer to MCMC reference, providing a viable path for trustworthy ABI across a range of different tasks.

Problem

• Amortized Bayesian Inference
• Unsupervised Continual Learning
• Model Misspecification
• Catastrophic Forgetting
• Self-Consistency Training

Innovation

• Continual Learning
• Amortized Bayesian Inference
• Self-Consistency
• Catastrophic Forgetting
• Elastic Weight Consolidation