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。

"Machine Learning Pipeline" as a mental map should be "considered harmful"

8 点作者 jamesblonde超过 1 年前

1 comment

jamesblonde超过 1 年前
Disclaimer - i am the author. This is a pun on Dijkstra&#x27;s &quot;Go To Statement Considered Harmful&quot;.<p>In 2023, the term “machine learning pipeline” is excessively used as a way &quot;to productionize ML models&quot;. However, there is widespread confusion about what a ML pipeline is - what its inputs and outputs are. A pipeline is a computer program that has clearly defined inputs and outputs and either runs on a schedule or is run continuously. The problem is that when you say to somebody “build a ML pipeline” you are not conveying any information apart from - you need to automate the execution of a program. The goal of this article is to say - be precise about what you are talking about.<p>Are you creating features? - it is a feature pipeline Are you training a model? - it is a training pipeline Are you making predictions with a model? - it is an inference pipeline