arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2103.14343math.OC
收藏

作为最优控制问题的神经网络训练:一种增广拉格朗日方法

Neural Network Training as an Optimal Control Problem: An Augmented Lagrangian Approach

Brecht Evens, Puya Latafat, Andreas Themelis, Johan Suykens, Panagiotis Patrinos

摘要:

神经网络的训练相当于非凸最优化问题,通常通过使用反向传播和随机梯度下降(的变体)来解决。在这项工作中,我们提出了另一种方法,通过查看训练任务作为一个非线性最优控制问题。在这个镜头下,反向传播相当于顺序方法(单次射击)的最优控制,其中的状态变量已被删除。众所周知,单次照射可能导致病态,因此通常优选同时进行的方法(多次照射)。出于这一假设,一个增广拉格朗日算法的开发,只需要一个近似解的拉格朗日子问题的用户定义的精度。通过将此框架应用于神经网络的训练,结果表明,内部拉格朗日子问题可以使用高斯-牛顿迭代来解决。为了充分利用神经网络的结构,由此产生的线性最小二乘问题,采用基于前向动态规划的方法来解决。最后,在回归数据集上展示了我们方法的有效性。

英文摘要:

Training of neural networks amounts to nonconvex optimization problems that are typically solved by using backpropagation and (variants of) stochastic gradient descent. In this work we propose an alternative approach by viewing the training task as a nonlinear optimal control problem. Under this lens, backpropagation amounts to the sequential approach (single shooting) to optimal control, where the states variables have been eliminated. It is well known that single shooting may lead to ill conditioning, and for this reason the simultaneous approach (multiple shooting) is typically preferred. Motivated by this hypothesis, an augmented Lagrangian algorithm is developed that only requires an approximate solution to the Lagrangian subproblems up to a user-defined accuracy. By applying this framework to the training of neural networks, it is shown that the inner Lagrangian subproblems are amenable to be solved using Gauss-Newton iterations. To fully exploit the structure of neural networks, the resulting linear least squares problems are addressed by employing an approach based on forward dynamic programming. Finally, the effectiveness of our method is showcased on regression datasets.

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。

↑