ReLU-KAN:仅需矩阵加法、逐点乘法和 ReLU 的新型 Kolmogorov-Arnold 网络
ReLU-KAN: New Kolmogorov-Arnold Networks that Only Need Matrix Addition, Dot Multiplication, and ReLU
AI总结:
针对 KAN 因 B 样条基函数导致 GPU 并行效率低的问题,本文提出仅用 ReLU 与逐点乘法的 ReLU-KAN,在 4 层网络上较传统 KAN 提速 20 倍,并保持更稳定训练、更强拟合及抗遗忘特性。
摘要:
受限于基函数(B 样条)计算的复杂性,Kolmogorov-Arnold Networks(KAN)在 GPU 上的并行计算能力受到限制。本文提出一种新颖的 ReLU-KAN 实现,它继承了 KAN 的核心思想。通过采用 ReLU(Rectified Linear Unit,修正线性单元)和逐点乘法,我们简化了 KAN 基函数的设计,并优化了计算流程,以实现高效的 CUDA 计算。所提出的 ReLU-KAN 架构可以在现有深度学习框架(例如 PyTorch)中方便地实现,用于推理和训练。实验结果表明,在 4 层网络上,ReLU-KAN 相比传统 KAN 实现了 20 倍加速。此外,ReLU-KAN 表现出更稳定的训练过程和更强的拟合能力,同时保留了 KAN 的“避免灾难性遗忘”特性。代码可在 https://github.com/quiqi/relu_kan 获取。
英文摘要:
Limited by the complexity of basis function (B-spline) calculations, Kolmogorov-Arnold Networks (KAN) suffer from restricted parallel computing capability on GPUs. This paper proposes a novel ReLU-KAN implementation that inherits the core idea of KAN. By adopting ReLU (Rectified Linear Unit) and point-wise multiplication, we simplify the design of KAN's basis function and optimize the computation process for efficient CUDA computing. The proposed ReLU-KAN architecture can be readily implemented on existing deep learning frameworks (e.g., PyTorch) for both inference and training. Experimental results demonstrate that ReLU-KAN achieves a 20x speedup compared to traditional KAN with 4-layer networks. Furthermore, ReLU-KAN exhibits a more stable training process with superior fitting ability while preserving the "catastrophic forgetting avoidance" property of KAN. You can get the code in https://github.com/quiqi/relu_kan
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。