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 2006.07928cs.LGcs.ITmath.ITmath.OCstat.ML
收藏

过参数神经网络Sobolev训练的全局收敛性

Global Convergence of Sobolev Training for Overparameterized Neural Networks

Northeastern University(东北大学)

Jorio Cocola, Paul Hand

摘要:

Sobolev损失用于训练网络以近似目标函数在指定输入点集合处的值和导数。最近的工作已经证明了它在各种任务中的成功应用,如蒸馏或合成梯度预测。在这项工作中,我们证明了一个过参数化的两层relu神经网络训练的Sobolev损失与梯度流从随机初始化可以适合任何给定的函数值和任何给定的方向导数,分离条件下的输入数据。

英文摘要:

Sobolev loss is used when training a network to approximate the values and derivatives of a target function at a prescribed set of input points. Recent works have demonstrated its successful applications in various tasks such as distillation or synthetic gradient prediction. In this work we prove that an overparameterized two-layer relu neural network trained on the Sobolev loss with gradient flow from random initialization can fit any given function values and any given directional derivatives, under a separation condition on the input data.

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。

↑