Skip to content
Sandeep's Quicktake: Getting teams to not just look at wrong labels in ML data episode artwork

Dec 11, 2021 2 min

Sandeep's Quicktake: Getting teams to not just look at wrong labels in ML data

from Journeys to Democratize Data+AI Hosted by Sandeep Uttamchandani

Democratize is not one big strategy but 100s of small things that you put in place. Quicktake is a series of short practical tips to get you started towards making data and AI widely accessible and self-service within your organization. This tip covers an important myth: To improve model accuracy, start by verifying the correctness of labels. Typically, there is only a small percentage of miss predictions that are related to wrong labels. Often times the biggest reason for model inaccuracies is the poor quality of data samples. Train the team just that instead of jumping to fix the incorrect labels, start by analyzing a sample of results that were misclassified and make a judgment call on whether to invest in fixing the labels going back and looking at opus deposit useful.

Episode details come from the podcast's RSS feed.

Embed this episode

Ready to play

Sandeep's Quicktake: Getting teams to not just look at wrong labels in ML data

0:00 2:52

No transcript yet.

Requests help us decide what to transcribe next.

Frequently asked questions

How long is this episode of Journeys to Democratize Data+AI?

This episode is 2 minutes long.

When was this Journeys to Democratize Data+AI episode published?

This episode was published on December 11, 2021.

Can I download this Journeys to Democratize Data+AI episode?

Yes. Use the download control on the episode player to save the publisher-provided media file.