Answers are grounded in this paper and cite their sources.
Theoretical analysis of low-rank adaptation reveals its effectiveness for fine-tuning models, suggesting broad implications for deep learning.
Low-Rank Adaptation (LoRA), a parameter-efficient fine-tuning method that leverages low-rank adaptation of weight matrices, has emerged as a prevalent technique for fine-tuning pre-trained models such as large language models and diffusion models. Despite its huge success in practice, the theoretical underpinnings of LoRA have largely remained unexplored. This paper takes the first step to bridge this gap by theoretically analyzing the expressive power of LoRA. We prove that, for fully connected neural networks, LoRA can adapt any model f to accurately represent any smaller target model f̄ if LoRA-rank ≥(width of f) × depth of f̄depth of f. We also quantify the approximation error when LoRA-rank is lower than the threshold. For Transformer networks, we show any model can be adapted to a target model of the same size with rank-(embedding size/2) LoRA adapters.
No takes yet. Share an insight, caveat, or question.
Zeng et al. (2023) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: