[edit]
LoRA Training in the NTK Regime has No Spurious Local Minima
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:21306-21328, 2024.
Abstract
Low-rank adaptation (LoRA) has become the standard approach for parameter-efficient fine-tuning of large language models (LLM), but our theoretical understanding of LoRA has been limited. In this work, we theoretically analyze LoRA fine-tuning in the neural tangent kernel (NTK) regime with N data points, showing: (i) full fine-tuning (without LoRA) admits a low-rank solution of rank r≲; (ii) using LoRA with rank r\gtrsim \sqrt{N} eliminates spurious local minima, allowing gradient descent to find the low-rank solutions; (iii) the low-rank solution found using LoRA generalizes well.