Generate to Discriminate: Expert Routing for Continual Learning

2024-12-22

AI Summary

In privacy-sensitive, cross-institutional settings where raw data sharing is prohibited, continual learning faces the dual challenge of mitigating catastrophic forgetting during domain expansion while enabling adaptive inference without access to historical data. Method: We propose a novel “Generate-for-Discrimination” paradigm: lightweight domain discriminators are trained exclusively on synthetic data to dynamically route test inputs to specialized experts, obviating the need for direct fine-tuning. Our approach unifies domain-incremental continual learning, synthetic-data-driven discriminative modeling, and multi-expert ensembling. Contribution/Results: Without accessing original data, our method simultaneously preserves performance on previously encountered domains and enables rapid adaptation to new ones. Evaluated on vision-language multi-task benchmarks, it significantly outperforms state-of-the-art domain-incremental methods. Notably, synthetic-data-trained discriminators achieve superior routing accuracy compared to those trained with real-label supervision—demonstrating that high-fidelity synthetic data can effectively substitute for private ground-truth labels in expert selection. This establishes a scalable, privacy-preserving framework for collaborative model deployment across institutions.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

In many real-world settings, regulations and economic incentives permit the sharing of models but not data across institutional boundaries. In such scenarios, practitioners might hope to adapt models to new domains, without losing performance on previous domains (so-called catastrophic forgetting). While any single model may struggle to achieve this goal, learning an ensemble of domain-specific experts offers the potential to adapt more closely to each individual institution. However, a core challenge in this context is determining which expert to deploy at test time. In this paper, we propose Generate to Discriminate (G2D), a domain-incremental continual learning method that leverages synthetic data to train a domain-discriminator that routes samples at inference time to the appropriate expert. Surprisingly, we find that leveraging synthetic data in this capacity is more effective than using the samples to extit{directly} train the downstream classifier (the more common approach to leveraging synthetic data in the lifelong learning literature). We observe that G2D outperforms competitive domain-incremental learning methods on tasks in both vision and language modalities, providing a new perspective on the use of synthetic data in the lifelong learning literature.

Problem

• Continual Learning
• Knowledge Retention
• Adaptive Model Selection

Innovation

• G2D Method
• Synthetic Data
• Continuous Learning