TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Why machine learning algorithms are hard to tune and how to fix it

1 点作者 brutuscat大约 4 年前

1 comment

brutuscat大约 4 年前
> Nothing of this article is novel. You might already be aware of everything we wanted to say. However, we have the impression that most machine learning curricula do not discuss optimisation methods very well (I know mine did not), and consequently, gradient descent is being treated as the one method to solve all problems. And the general message is that if an algorithm does not work for your problem, you need to spend more time tuning the hyper-parameters to your problem.