arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2002.01523cs.LGstat.ML
收藏

神经网络的深度条件化处理

A Deep Conditioning Treatment of Neural Networks

Google AI(谷歌人工智能);Princeton(普林斯顿大学);Google Research(谷歌研究院)

Naman Agarwal, Pranjal Awasthi, Satyen Kale

摘要:

我们研究了深度在训练随机初始化的超参数化神经网络中的作用。我们给出了一个一般性的结果,表明深度通过改善输入数据的某些核矩阵的条件来提高神经网络的可训练性。这一结果适用于任意非线性激活函数在一定的归一化。我们提供了结果的版本,通过神经切线核只训练神经网络的顶层,以及训练所有层。作为这些一般结果的应用,我们提供了Das等人(2019)结果的推广,表明具有大量非线性激活的深度随机神经网络的可学习性随深度呈指数级下降。我们还通过Bartlett et al.(2019 b)的结果展示了良性过拟合如何在深度神经网络中发生。我们还提供了实验证据,证明ReLU的规范化版本是训练深度神经网络中更复杂操作(如批量规范化)的可行替代方案。

英文摘要:

We study the role of depth in training randomly initialized overparameterized neural networks. We give a general result showing that depth improves trainability of neural networks by improving the conditioning of certain kernel matrices of the input data. This result holds for arbitrary non-linear activation functions under a certain normalization. We provide versions of the result that hold for training just the top layer of the neural network, as well as for training all layers, via the neural tangent kernel. As applications of these general results, we provide a generalization of the results of Das et al. (2019) showing that learnability of deep random neural networks with a large class of non-linear activations degrades exponentially with depth. We also show how benign overfitting can occur in deep neural networks via the results of Bartlett et al. (2019b). We also give experimental evidence that normalized versions of ReLU are a viable alternative to more complex operations like Batch Normalization in training deep neural networks.

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。

↑