【源码】L1正则化的最优化方法

L1正则化的最优化方法

在本文中，我们回顾并比较了用于解决服从“L1-正则化”的二次可微损失函数最小化问题的最新优化技术。

In this paper we review and comparestate-of-the-art optimization techniques for solving the problem of minimizinga twice-differentiable loss function subject to L1-regularization.

本文工作的第一部分概述了大量可行的方法用于解决该类问题，并重点阐述了这些方法的优缺点。

The first part of this work outlines avariety of the approaches that are available to solve this type of problem,highlighting some of their strengths and weaknesses.

在第二部分中，我们比较了在不同场景下14种优化策略的数值结果。

In the second part, we present numericalresults comparing 14 optimization strategies under various scenarios.

我们实验评估的一个局限性是，虽然我们从功能上测量了评估方法的性能，用以作为实时运行时的比较，但这种比较很大程度上独立于底层的实现细节和实际的实验硬件结构体系。

A limitation of our empirical evaluation isthat we have measured the performance of the methods in terms of functionevaluations, which we used as a surrogate for runtime comparisons in order tomake our comparisons largely independent of low-level implementation detailsand of the actual hardware architecture used in performing the experiments.

与本文相关的网站供参考：

https://www.cs.ubc.ca/~schmidtm/Software/L1General.html

https://www.cs.ubc.ca/~schmidtm/Software/L1General/examples.html

下载英文原文及源码地址：

http://page2.dfpan.com/fs/6lccaj72d2a1e219166/

更多精彩文章请关注微信号：