Description:

Today in our daily lives, we use many products which use models trained using different machine learning (ML) techniques. There are many products being built which use the models at its core, such as driverless cars, search/recommendation engines, etc. To promote the growth of ML, build better, impartial and safer products, it is becoming crucial to fully understand internal processes of its algorithms. Inclusion of "right to explanation" for an output of the algorithm in GDPR signifies the growing need of transparency in ML models. Various visualization techniques have been shown to be very effective in achieving this goal. In this seminar, we offer various topics which explore different aspects of ML models with the help of visualization.

Possible topics:

1. Visualizations of computer game AI agents

2. Visual analytics for image classification algorithms

3. Visual analytics for sequence models

4. Visual investigation in text analytics

5. Visual analytics for multiclass classifiers

6. Visualization techniques to show uncertainty in machine learning models

If you consider participating, please attend the first session, Wednesday, April 10. We'll present details and assign topics then.

Learning Targets: