arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2201.12250cs.LG
收藏

神经元上的梯度下降及其与近似二阶优化的联系

Gradient Descent on Neurons and its Link to Approximate Second-Order Optimization

ETH Zurich(苏黎世联邦理工学院)

Frederik Benzing

摘要:

二阶优化器被认为具有加速神经网络训练的潜力,但由于曲率矩阵的巨大尺寸,它们通常需要近似值在计算上易于处理。最成功的近似族是Kronecker因子,块对角曲率估计(KFAC)。在这里,我们结合了以前的工作,以评估精确的二阶更新与仔细的消融,以建立一个令人惊讶的结果:由于其近似,KFAC是不密切相关的二阶更新,特别是,它显着优于真正的二阶更新。这对人们普遍持有的信念提出了挑战,并立即提出了一个问题,即为什么KFAC表现如此出色。为了回答这个问题,我们提出的证据强烈表明,KFAC近似一阶算法,它执行梯度下降的神经元,而不是权重。最后,我们表明,这种优化器往往提高了KFAC的计算成本和数据效率。

英文摘要:

Second-order optimizers are thought to hold the potential to speed up neural network training, but due to the enormous size of the curvature matrix, they typically require approximations to be computationally tractable. The most successful family of approximations are Kronecker-Factored, block-diagonal curvature estimates (KFAC). Here, we combine tools from prior work to evaluate exact second-order updates with careful ablations to establish a surprising result: Due to its approximations, KFAC is not closely related to second-order updates, and in particular, it significantly outperforms true second-order updates. This challenges widely held believes and immediately raises the question why KFAC performs so well. Towards answering this question we present evidence strongly suggesting that KFAC approximates a first-order algorithm, which performs gradient descent on neurons rather than weights. Finally, we show that this optimizer often improves over KFAC in terms of computational cost and data-efficiency.

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。

↑