论文部分内容阅读
一般说来,训练深度学习网络的方式主要有四种:监督、无监督、半监督和强化学习。本文编译自硅谷著名的风险投资机构安德森·霍洛维茨基金,接下来,作者Frank Chen将逐个解释这些方法背后所蕴含的理论知识。除此之外,还将分享文献中经常碰到的术语,并提供与数学相关的更多资源。有关数学相关问题,请参阅斯坦福大学的这个教程,其中包含监督和无监督学习,内含代码示例。
In general, there are mainly four ways to train a deep learning network: supervision, unsupervised, semi-supervised and intensive learning. This article is compiled from the Anderson Horowitz Fund, a well-known venture capital firm in Silicon Valley. Next, the author Frank Chen will explain one by one the theoretical knowledge behind these methods. In addition, terms often encountered in the literature will be shared and more resources related to mathematics will be provided. For math-related questions, see this tutorial from Stanford University, which includes supervised and unsupervised learning with included code examples.