BriefGPT.xyz
Oct, 2016
应用非线性谱方法全局最优训练广义多项式神经网络
Globally Optimal Training of Generalized Polynomial Neural Networks with Nonlinear Spectral Methods
HTML
PDF
Antoine Gautier, Quynh Nguyen, Matthias Hein
TL;DR
本研究提出了一种非线性谱方法用于全局优化某类前馈神经网络，在弱数据假设下可以实现全局最优解，而不需要手动参数调整，这是首个实际可行的具有全局最优性保证的方法。
Abstract
The
optimization
problem behind
neural networks
is highly non-convex. Training with stochastic gradient descent and variants requires careful
par
→