This is a Self-Organizing Map, trained on sources from the LOFAR survey. Click on one of these prototypes.
On the left you can see where the radio source you clicked on is located on the sky. The source might be accompanied or interacting with other sources or be part of some larger structure!

About this project

From the shape or morphology of a radio source we can infer physical properties of the source and its environment.

To find out what different morphologies are present in the LOFAR radio survey, we use a dimensionality reduction technique known as a Self-Organizing Map.

This is an unsupervised neural network that projects a high-dimensional dataset to a discrete 2-dimensional representation.

The map contains 10 x 10 neurons or prototypes, each represents a cluster of sources.

The radio data we used, with frequencies between 120 and 168Mhz, is part of the LoTSS wide area survey.