Proxy-informed Bayesian transfer learning with unknown sources

2024-11-05

arXiv.org

AI Summary

This paper addresses negative transfer under unknown source domains—where transferring source knowledge degrades target-domain performance—by proposing PROMPT, the first Bayesian causal framework for this setting. Under stringent conditions—no target-domain labels, no source-domain priors, and reliance solely on noisy proxy signals (e.g., human feedback)—PROMPT models the transferability of causal mechanisms and structural uncertainty to distinguish transferable from non-transferable causal paths. It innovatively formalizes the root cause of negative transfer from a Bayesian perspective and theoretically proves that its risk is independent of the proxy signal’s signal-to-noise ratio. PROMPT requires neither source-knowledge fine-tuning nor explicit access to source data, and supports cross-task transfer with latent variables. Experiments demonstrate that it significantly mitigates negative transfer in settings with implicit sources and sparse feedback, substantially improving target-domain generalization.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Reasoning under Uncertainty: Causality
• Natural Language Processing: Prompt Engineering / Prompting

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Generalization outside the scope of one's training data requires leveraging prior knowledge about the effects that transfer, and the effects that don't, between different data sources. Transfer learning is a framework for specifying and refining this knowledge about sets of source (training) and target (prediction) data. A challenging open problem is addressing the empirical phenomenon of negative transfer, whereby the transfer learner performs worse on the target data after taking the source data into account than before. We first introduce a Bayesian perspective on negative transfer, and then a method to address it. The key insight from our formulation is that negative transfer can stem from misspecified prior information about non-transferable causes of the source data. Our proposed method, proxy-informed robust method for probabilistic transfer learning (PROMPT), does not require prior knowledge of the source data (the data sources may be"unknown"). PROMPT is thus applicable when differences between tasks are unobserved, such as in the presence of latent confounders. Moreover, the learner need not have access to observations in the target task (cannot"fine-tune"), and instead makes use of proxy (indirect) information. Our theoretical results show that the threat of negative transfer does not depend on the informativeness of the proxy information, highlighting the usefulness of PROMPT in cases where only noisy indirect information, such as human feedback, is available.

Problem

• Addressing negative transfer in transfer learning
• Leveraging proxy data without known sources
• Handling latent confounders in task differences

Innovation

• Bayesian transfer learning method
• Proxy-informed robust approach
• No need for source data knowledge