The Quest for Winning Tickets in Low-Rank Adapters
By: Hamed Damirchi , Cristian Rodriguez-Opazo , Ehsan Abbasnejad and more
The Lottery Ticket Hypothesis (LTH) suggests that over-parameterized neural networks contain sparse subnetworks ("winning tickets") capable of matching full model performance when trained from scratch. With the growing reliance on fine-tuning large pretrained models, we investigate whether LTH extends to parameter-efficient fine-tuning (PEFT), specifically focusing on Low-Rank Adaptation (LoRA) methods. Our key finding is that LTH holds within LoRAs, revealing sparse subnetworks that can match the performance of dense adapters. In particular, we find that the effectiveness of sparse subnetworks depends more on how much sparsity is applied in each layer than on the exact weights included in the subnetwork. Building on this insight, we propose Partial-LoRA, a method that systematically identifies said subnetworks and trains sparse low-rank adapters aligned with task-relevant subspaces of the pre-trained model. Experiments across 8 vision and 12 language tasks in both single-task and multi-task settings show that Partial-LoRA reduces the number of trainable parameters by up to 87\%, while maintaining or improving accuracy. Our results not only deepen our theoretical understanding of transfer learning and the interplay between pretraining and fine-tuning but also open new avenues for developing more efficient adaptation strategies.
Similar Papers
Towards Higher Effective Rank in Parameter-efficient Fine-tuning using Khatri--Rao Product
Machine Learning (CS)
Makes AI learn better without needing more power.
LoFT: Low-Rank Adaptation That Behaves Like Full Fine-Tuning
Machine Learning (CS)
Makes smart computer models learn better and faster.
DropLoRA: Sparse Low-Rank Adaptation for Parameter-Efficient Fine-Tuning
Computation and Language
Makes AI smarter without more training.