technosignatures
Breakthrough Listen: Deep Learning Search for Technosignatureshttps://seti.berkeley.edu/ml_gbt/Machine learning algorithms applied to Breakthrough Listen data from the Green Bank Telescope have been used to find signals of interest, while filtering out radio frequency interference, as reported in a new paper accepted for publication in Nature Astronomy. Follow-up observations did not re-detect the signals, so they do not pass the criteria required for bona fide technosignature candidates. However, the technique represents a promising new way to look for anomalies in data from our searches, while efficiently filtering out millions of signals from our own technology.
A paper describing these results (see also shareable PDF link) has been accepted for publication in the journal Nature Astronomy and a preprint is also available.
Will Machine Learning Help Us Find Extraterrestrial Life?https://www.seti.org/press-release/will-machine-learning-help-us-find-extraterrestrial-lifeThis study re-examined data taken with the Green Bank Telescope in West Virginia as part of a Breakthrough Listen campaign that initially indicated no targets of interest. The goal was to apply new deep learning techniques to a classical search algorithm to yield faster, more accurate results. After running the new algorithm and manually re-examining the data to confirm the results, newly detected signals had several key characteristics:
The signals were narrow band, meaning they had narrow spectral width, on the order of just a few Hz. Signals caused by natural phenomena tend to be broadband.
- The signals had non-zero drift rates, which means the signals had a slope. Such slopes could indicate a signal’s origin had some relative acceleration with our receivers, hence not local to the radio observatory.
- The signals appeared in ON-source observations and not in OFF-source observations. If a signal originates from a specific celestial source, it appears when we point our telescope toward the target and disappears when we look away. Human radio interference usually occurs in ON and OFF observations due to the source being close by.
- Cherry Ng, another of Ma’s research advisors and an astronomer at both the SETI Institute and the French National Center for Scientific Research said, “These results dramatically illustrate the power of applying modern machine learning and computer vision methods to data challenges in astronomy, resulting in both new detections and higher performance. Application of these techniques at scale will be transformational for radio technosignature science.”
While re-examinations of these new targets of interest have yet to result in re-detections of these signals, this new approach to analyzing data can enable researchers to more effectively understand the data they collect and act quickly to re-examine targets.