arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2102.04374cs.ITmath.IT
收藏

神经网络初始化的互信息:平均场近似

Mutual Information of Neural Network Initialisations: Mean Field Approximations

Jared Tanner, Giuseppe Ughi

摘要:

已知训练随机初始化的深度神经网络的能力强烈依赖于权重矩阵和偏置的方差以及非线性激活的选择。在这里,我们补充现有的几何分析这种现象与信息理论的替代。下界推导出输入和隐藏层输出之间的互信息。使用平均场分析,我们能够提供分析下限作为网络权重和偏差方差的函数,以及非线性激活的选择。这些结果表明,从训练的角度来看,已知是最佳的初始化从互信息的角度来看也是优越的。

英文摘要:

The ability to train randomly initialised deep neural networks is known to depend strongly on the variance of the weight matrices and biases as well as the choice of nonlinear activation. Here we complement the existing geometric analysis of this phenomenon with an information theoretic alternative. Lower bounds are derived for the mutual information between an input and hidden layer outputs. Using a mean field analysis we are able to provide analytic lower bounds as functions of network weight and bias variances as well as the choice of nonlinear activation. These results show that initialisations known to be optimal from a training point of view are also superior from a mutual information perspective.

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。

↑