Dominos as an analogy for dependencies, pixabay.com.

Member-only story

Monitor Your Dependencies! Stop Being A Blind Data-Scientist.

Reasons for monitoring your model dependencies.

Dr. Ori Cohen
4 min readApr 4, 2020

--

In my previous article “Monitor! Stop Being A Blind Data Scientist”, I mentioned the many use cases and the monumental importance of monitoring & alerts on our field, specifically from a data-science-researcher point-of-view. I looked at use-cases and reviewed several companies that provide varying solutions for this huge, but strangely undiscussed problem, i.e., the lack of model monitoring, a problem that atm overruns our industry.

Update: this article has now been published as article #49 in O’reilly’s book “97 things every cloud engineer should know”

Use Cases

In the original article, I listed several use cases for data monitoring & alerts, I am sure there are others, but these ones that have the most impact on a data scientist. The list was written in chronological order. In this article, we’ll look into package dependency-versions and explore several incidents that can affect your clients directly.

  1. Annotators Performance
  2. Annotation Distribution
  3. Data Integrity, i.e., schemas and test-driven data analysis.
  4. Data Distributions, i.e., concept drifts

--

--

Towards Data Science
Towards Data Science

Published in Towards Data Science

Your home for data science and AI. The world’s leading publication for data science, data analytics, data engineering, machine learning, and artificial intelligence professionals.

No responses yet