arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2201.05125cs.LGcs.CV
收藏

GradMax:使用梯度信息生长神经网络

GradMax: Growing Neural Networks using Gradient Information

Google Research(谷歌研究院)

Utku Evci, Bart van Merriënboer, Thomas Unterthiner, Max Vladymyrov, Fabian Pedregosa

摘要:

神经网络的结构和参数通常是独立优化的,这需要在结构被修改时对参数进行昂贵的重新训练。在这项工作中,我们专注于增长架构,而不需要昂贵的再培训。我们提出了一种方法,在训练过程中添加新的神经元,而不影响已经学到的东西,同时提高训练动态。我们通过最大化新权重的梯度来实现后者,并通过奇异值分解(SVD)有效地找到最佳初始化。我们称这种技术为梯度最大化增长(GradMax),并在各种视觉任务和架构中证明了其有效性。

英文摘要:

The architecture and the parameters of neural networks are often optimized independently, which requires costly retraining of the parameters whenever the architecture is modified. In this work we instead focus on growing the architecture without requiring costly retraining. We present a method that adds new neurons during training without impacting what is already learned, while improving the training dynamics. We achieve the latter by maximizing the gradients of the new weights and find the optimal initialization efficiently by means of the singular value decomposition (SVD). We call this technique Gradient Maximizing Growth (GradMax) and demonstrate its effectiveness in variety of vision tasks and architectures.

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。

↑