On Vanishing Gradients, Over-Smoothing, and Over-Squashing in GNNs: Bridging Recurrent and Graph Learning
"梯度消失" 相关文章
From GAN to WGAN
非线性激活
The Vanishing Gradient Problem for Stiff Neural Differential Equations
Continuous Spiking Graph Neural Networks
用更好的方式来监控神经网络的训练过程
【漫话机器学习系列】185.神经网络参数的标准初始化（Normalized Initialization of Neural Network Parameter
University of South Florida Researchers Propose TeLU Activation Function for Fast and Stable Deep Learning
字节豆包大模型团队突破残差连接局限！预训练收敛最快加速80%
