无回放的连续低秩适应与动态记忆
Replay-Free Continual Low-Rank Adaptation with Dynamic Memory
AI总结：
本文提出DualLoRA方法，通过引入正交和残差低秩适配器，结合动态记忆机制，在连续学习中提升稳定性与可塑性，实现更高的准确率、推理速度和计算效率。
摘要：
我们重新审视连续学习（CL），使预训练的视觉变换器（ViTs）能够随时间顺序微调新的下游任务。然而，随着模型规模的增加，灾难性遗忘成为更严重的挑战。最近的研究强调了CL技术与参数高效微调（PEFT）之间的交叉，后者专注于仅微调少量可训练参数以适应下游任务，如低秩适应（LoRA）。虽然LoRA实现更快的收敛速度和更少的可训练参数，但很少在连续学习的背景下被探索。为解决这一差距，我们提出了一种新的PEFT-CL方法，称为双低秩适应（DualLoRA），在每个层中引入正交LoRA适配器和残差LoRA适配器，与预训练权重并行。这些组件由动态记忆机制协调，以在稳定性和可塑性之间取得平衡。此外，我们提出了一种预测任务身份并具有信心的方案，并校准模型的输出。在基于ViT的模型上，我们证明DualLoRA在多个基准测试中在准确性、推理速度和训练计算效率方面具有显著优势。
英文摘要：
We revisit continual learning~(CL), which enables pre-trained vision transformers (ViTs) to sequentially fine-tune on new downstream tasks over time. However, as the scale of these models increases, catastrophic forgetting remains a more serious challenge. Recent studies highlight a crossover between CL techniques and parameter-efficient fine-tuning (PEFT), which focuses on fine-tuning only a small set of trainable parameters to adapt to downstream tasks, such as low-rank adaptation (LoRA). While LoRA achieves faster convergence and requires fewer trainable parameters, it has seldom been explored in the context of continual learning. To address this gap, we propose a novel PEFT-CL method called Dual Low-Rank Adaptation (DualLoRA), which introduces both an orthogonal LoRA adapter and a residual LoRA adapter parallel to pre-trained weights in each layer. These components are orchestrated by a dynamic memory mechanism to strike a balance between stability and plasticity. Additionally, we propose a scheme to predict task identity with confidence and calibrate the model's outputs accordingly. On ViT-based models, we demonstrate that DualLoRA offers significant advantages in accuracy, inference speed, and computation efficiency in training over existing CL methods across multiple benchmarks.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
