Case Study: NASA achieves 3x improvement in gravitational lensing accuracy with Kaggle

NASA - Customer Case Study

Mapping Dark Matter was a research competition sponsored by NASA, the Royal Astronomical Society and the European Space Agency to tackle a key astronomy challenge: how to correct for invisible “dark matter” that distorts light from distant galaxies. Competitors were given 100,000 PNG images of blurred, simulated galaxies and had three months to recover the true galaxy shapes (e1 and e2 values) so gravitational lensing could be measured accurately.

Teams from diverse fields submitted 760 entries across about 70 teams; the winning approaches used artificial neural networks (notably work from UC Irvine and others) and produced a threefold increase in accuracy over the previous state of the art. The improvement is significant enough that the methods are likely to be used in ESA’s Euclid space telescope, and winners received an expenses-paid trip to present their work at NASA JPL.

NASA

Tom Kitching

The Royal Astronomical Society