Learning to Learn Weight Generation via Local Consistency Diffusion
This addresses the problem of efficient and generalizable weight generation for machine learning practitioners, though it appears incremental as it builds on existing diffusion and meta-learning techniques.
The paper tackles the limitations of diffusion-based weight generation methods in generalizability and local target assignment by proposing Mc-Di, which integrates meta-learning with a local consistency diffusion algorithm. Experiments show superior accuracy and inference efficiency in tasks like transfer learning and large language model adaptation.
Diffusion-based algorithms have emerged as promising techniques for weight generation. However, existing solutions are limited by two challenges: generalizability and local target assignment. The former arises from the inherent lack of cross-task transferability in existing single-level optimization methods, limiting the model's performance on new tasks. The latter lies in existing research modeling only global optimal weights, neglecting the supervision signals in local target weights. Moreover, naively assigning local target weights causes local-global inconsistency. To address these issues, we propose Mc-Di, which integrates the diffusion algorithm with meta-learning for better generalizability. Furthermore, we extend the vanilla diffusion into a local consistency diffusion algorithm. Our theory and experiments demonstrate that it can learn from local targets while maintaining consistency with the global optima. We validate Mc-Di's superior accuracy and inference efficiency in tasks that require frequent weight updates, including transfer learning, few-shot learning, domain generalization, and large language model adaptation.