
Using artificial intelligence to create detailed maps of galaxy structures and dark matter distribution.Neural network galaxy mapping uses artificial intelligence to create detailed maps of galaxy structures and dark matter distribution. This approach has become essential because dark matter, which makes up most of the universe's mass, does not emit light and can only be inferred through its gravitational effects. By training neural networks on large datasets of galaxy images and simulations, researchers can identify patterns that reveal the underlying dark matter scaffolding, including filaments, halos, and voids.
The application of neural networks to galaxy mapping gained traction in the 2010s, with major advances driven by deep learning and large surveys like SDSS and DES. The technique became widely adopted after successes in classifying galaxies and detecting gravitational lenses.
Neural networks, particularly convolutional and graph-based architectures, learn to recognize features in galaxy survey data. They are trained on simulated universes where the true dark matter distribution is known, then applied to real observations. The networks can detect subtle distortions in galaxy shapes caused by gravitational lensing, or classify galaxies by morphology to trace large-scale structure, effectively filling in gaps where direct measurement is impossible.
Used in cosmological surveys to create dark matter maps, identify gravitational lenses, and improve estimates of cosmic parameters like the matter density and clustering amplitude.
A neural network once discovered 247 new gravitational lenses in a single survey, more than many human-led searches had found in years.
























