About the library
How the visualisations are made.
Suranjan Poudel builds and maintains this collection of interactive machine-learning visualisations. Each page covers one topic, includes its assumptions and references, and runs without accounts or visitor analytics.
Approach
Each exhibit focuses on one relationship. Controls change the model or data directly, and guided steps point out useful states without requiring a fixed order. Fixed models and datasets keep the examples repeatable.
Evidence and simplification
The insight drawer for every exhibit names its references and states what is illustrative, idealised, or omitted. Equations appear only when they clarify a control or visible mechanism. The visualisations are explanations, not production model benchmarks or claims about real-world performance.
Accessibility
Controls are designed for keyboard, pointer, and touch input. Scenes include a nonvisual description, important states are not encoded by colour alone, dialogs trap and restore focus, and automatic walkthroughs stop when reduced motion is preferred. Automated checks complement—not replace—manual review with assistive technology and real devices.
Privacy
The site is a static export. It uses no analytics, tracking pixels, cookies, accounts, advertising, or stored visitor profiles. Search, filters, and exhibit state remain in the browser and are represented in ordinary URLs when they need to be shared.
Source, licence, and reuse
The source is publicly available on GitHub. The repository does not currently include an explicit software licence, so source availability should not be interpreted as permission to redistribute it; reuse terms will be documented there if that changes.
View source repositoryShare or embed an exhibit
Use “Copy exhibit link” in any workspace to copy its normal page address. For a navigation-free iframe, add ?embed=1 to that address. Embedded views retain keyboard controls and a link back to the full page.