Concept Drift Detection with NannyML | Webinar

Wojtek introduces NannyML’s latest algorithm fully capable of detecting concept drift and estimating its magnitude.

After watching you will understand:
The 2 reasons why machine learning models fail
How to estimate concept drift magnitude
How to automatically retrain your models with concept drift detection

About NannyML Cloud ☁️
https://www.nannyml.com/nannyml-cloud

It’s available on the Azure and AWS Marketplaces, nannyML cloud is a complete tool to monitor post-deployment model performance (without access to targets). It’s free to get started.

Data Scientists can rely on nannyML cloud to detect and alert them ONLY when changes in data impact model performance. With advanced algorithms to estimate model performance, Data Scientists can spend more time focusing on what matters: optimizing model performance and ensuring their models drive business impact.

nannyML cloud empowers data scientists with advanced, easy-to-use:

Performance Monitoring Tools without access to targets
Root Cause Analysis tools: concept drift detection, covariate drift detection (multivariate and univariate)
Easily deploy and integrate a monitoring system within their cloud privately and securely.

Timestamps:
0:00 Introduction
01:04 Agenda
02:35 ML Monitoring Flow
04:48 Why is performance a crucial step in monitoring?
05:56 Covariate shift and concept drift Explained
9:37 Covariate shift and concept drift
13:48 Reverse Concept Drift (RCD)
15:49 RCD - assumptions
18:10 Intuition behind RCD
20:47 RCD plot
21:45 Calculating the magnitude of concept drift
26:20 Performance impact estimation
32:10 RCD results
34:11 Impact of covariate shift
37:04 Concept drift and retraining
42:00 Conclusions