论文标题
基于共轭级别的ADAM用于随机优化及其在深度学习中的应用
Conjugate-gradient-based Adam for stochastic optimization and its application to deep learning
论文作者
论文摘要
本文提出了一种基于共轭分子的ADAM算法，将ADAM与非线性缀合物梯度方法融合在一起，并显示其收敛分析。关于文本分类和图像分类的数值实验表明，所提出的算法可以比现有的自适应随机优化算法训练更少的时期神经网络模型。
This paper proposes a conjugate-gradient-based Adam algorithm blending Adam with nonlinear conjugate gradient methods and shows its convergence analysis. Numerical experiments on text classification and image classification show that the proposed algorithm can train deep neural network models in fewer epochs than the existing adaptive stochastic optimization algorithms can.
