Bayesian Meta-Learning with Expert Feedback for Task-Shift Adaptation through Causal Embeddings
This addresses the challenge of task-shift adaptation in meta-learning for scenarios like clinical prediction, though it appears incremental by building on existing causal and Bayesian approaches.
The paper tackles the problem of meta-learning failing to adapt to out-of-distribution tasks due to negative transfer, by proposing a Bayesian meta-learning method that uses causal embeddings and expert feedback to improve transfer based on mechanistic similarity, resulting in reduced negative transfer and better adaptation in simulations and a clinical prediction setting.
Meta-learning methods perform well on new within-distribution tasks but often fail when adapting to out-of-distribution target tasks, where transfer from source tasks can induce negative transfer. We propose a causally-aware Bayesian meta-learning method, by conditioning task-specific priors on precomputed latent causal task embeddings, enabling transfer based on mechanistic similarity rather than spurious correlations. Our approach explicitly considers realistic deployment settings where access to target-task data is limited, and adaptation relies on noisy (expert-provided) pairwise judgments of causal similarity between source and target tasks. We provide a theoretical analysis showing that conditioning on causal embeddings controls prior mismatch and mitigates negative transfer under task shift. Empirically, we demonstrate reductions in negative transfer and improved out-of-distribution adaptation in both controlled simulations and a large-scale real-world clinical prediction setting for cross-disease transfer, where causal embeddings align with underlying clinical mechanisms.