Alibi Detect: an open source Python library focused on outlier, adversarial and drift detection. The package aims to cover both online and offline detectors for tabular data, text, images and time series.
00:00 Intro and a few words about Ashley
01:00 Alibi Detect works with Seldon Core but also as a standalone library!
02:23 The Alibi Detect GitHub repo and docs
05:10 Why detect drift? Responsible AI and the EU AI act
06:30 What is drift?
08:49 How do drift detectors work in Alibi Detect?
10:05 Deciding if we are observing drift with p-values...
12:55 A simple demo!
15:01 Creating some example data
17:23 Creating a drift detector
17:50 Running the drift detector
19:50 Drift detection on real data!
20:59 Another demo! Drift detection on image data
23:09 Extracting features to perform drift detection on
25:54 Creating a drift detector with a preprocessor
27:22 Running the drift detector on a GPU
29:51 Any advice? Monitor and explain your ML systems to increase confidence!
Links:
- GitHub Repo: github.com/SeldonIO/alibi-detect
- An overview of drift detection with Alibi Detect: docs.seldon.io/projects/alibi...
- "A hands-on introduction to drift detection" webinar: www.seldon.io/webinar/intro-t...
Free MLOps course: github.com/DataTalksClub/mlop...
Join DataTalks.Club: datatalks.club/slack.html
Our events: datatalks.club/events.html
Негізгі бет Ойын-сауық Open-Source Spotlight - Alibi Detect - Ashley Scillitoe
Пікірлер: 1