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。

Data Scientists Should Be Able to Deploy and Iterate Their Own Models

25 点作者 mikeyanderson大约 6 年前

1 comment

hadsed大约 6 年前
I fully agree, but the tooling is super immature. I think there&#x27;s going to be incredible opportunity for engineers to build tools for doing ML in a very efficient and scientifically rigorous way.<p>For instance, Jupyter is the best thing we have to an IDE for science. It is an incredibly innovative project, but it is not what we need. We need a Photoshop, a Visual Studio, a Final Cut Pro for doing ML.<p>There are a lot of interesting projects out there solving some of these problems. My favorite ones are Prodigy (by Explosion AI), Pachyderm, Paperspace to name a few. But it&#x27;ll be a decade I think until we get to a serious place with it as an industry.<p>I myself have found the process of understanding models after training is incredibly difficult. I&#x27;m talking analyzing misclassifications, visualizing embeddings, and looking at saliency maps. We just don&#x27;t know enough about how models work, and when we do it&#x27;s only after great effort that most small shops don&#x27;t have the resources for. This was true when I was trying to get my last company off the ground and is still true now that I&#x27;m running ML at my current company. There is a pretty big opportunity especially given that most cloud ML companies seem to focus on just training and deployment. Thinking about trying to start this myself actually given how much of a pain point it is for me today.
评论 #19848242 未加载
评论 #19850980 未加载