Model maintenance based on domain adaptation and transfer learning

Course Filter

Home Courses Model maintenance based on domain adaptation and transfer learning

Type of course:

Digital learning, Path

Language:

EN

Duration:

75 minutes

Workload:

4 hours

Proficiency:

Beginner

Target:

Professionals, Students

This learning path explores the definition of dataset shifts, types of prediction shifts, and how they relate within a framework of model maintenance. It explains the distinction between causes and types of model performance degradation, the definition of shift detection metrics, and the different solutions for model correction when shifts are detected.

Learning outcomes

The learner will be able to differentiate the types and causes for model degradation
The learner will be able to quantify types and causes of model degradation and interpret the monitoring metrics
The learner will be able to implement the methods for model update with unlabelled, partially, and totally labeled information

LessonDataset shifts for prediction performance degradation in machine learning

Course Content

Expand All

Dataset shifts for prediction performance degradation in machine learning 17 Unities

Expand

Single module Content

0% Complete 0/17 Steps

Dataset shifts for prediction performance degradation in machine learning

Supervised machine learning models

Model maintenance

What happens after deployment?

Performance degradation path

Concepts for model performance degradation

Causes: Covariate shift and prior shift

Causes: conditional shift

Causes: regression example

From causes to types

Types of prediction shifts

Types of prediction shifts

Causes and types: Application

Causes and types: Application

LessonDetecting dataset and prediction shifts for machine learning models

Course Content

Expand All

Detecting dataset and prediction shifts for machine learning models 13 Unities

Expand

Single module Content

0% Complete 0/13 Steps

Detecting dataset and prediction shifts for machine learning models

Overview of dataset and prediction shifts

What are the tools to detect dataset shifts and types of prediction shifts?

The regression model for this case-study

Metrics for shifts detection

Covariate shift 1

Covariate shift 2: In combination with conditional shift

Covariate shift 3: reflected in a bias shift of the predictions

Further reading

LessonMachine learning model maintenance after shift detection

Course Content

Expand All

Machine learning model maintenance after shift detection 18 Unities

Expand

Single module Content

0% Complete 0/18 Steps

Machine learning model maintenance after shift detection

Summary of dataset and prediction shifts

Roadmap of model maintenance

Zooming into model adjustment

Model building vs. Model adjusment

Single correction

Recalculation of the model

Transfer learning / Domain adaptation

Transfer learning / Domain adaptation methods

Domain Invariant Partial Least Squares (di-PLS)

Domain Adaptative Partial Least Squares (da-PLS)

Kernel Domain Adaptative Partial Least Squares (kda-PLS)

When is unsupervised and (semi) supervised model adjustment possible?

Further reading

LessonMachine learning model maintenance - Quiz

Course Content

Expand All

Machine learning model maintenance - Quiz 4 Unities

Expand

Single module Content

0% Complete 0/4 Steps

Machine learning model maintenance - Quiz

Machine learning model maintenance - Quiz

Topics

Digital Transformation, Data mining, Data Analytics

Provided by

Software Competence Center Hagenberg GmbH

Content created in 2023

+228 enrolled

49 €

Take this Course

Take the next step toward your learning goals

Course Includes

1 Quiz
1 Certificate of achievement

Related