Topcoder Challenge Finds Two New Comets For NASA
Topcoder members recently participated in a data science challenge conducted by NASA. The Comet Detection Marathon Match Challenge, conducted on behalf of NASA SOHO/ESA, aimed to develop artificial intelligence and machine learning (AI/ML) algorithms capable of detecting comets that reflect minimal light. These “category C” comments are very dim—and hard to detect with the human eye. Enter Topcoder's Data Science community.
ABOUT THE NASA SOHO COMET SEARCH CHALLENGE
Using Solar Heliophysics Observatory (SOHO) mission open-science data, members created novel AI/ML approaches to find previously overlooked comets with as few false positives as possible. The data set members used came from SOHO’s C2 telescope and comprised approximately 36,000 images spread across 2,950 comet observations. The Topcoder challenge ran for four weeks, drew 596 registrants from 73 countries, and found two previously undetected comets!
The winning algorithms will be utilized by NASA’s Solar and Heliospheric Observatory (SOHO) satellite to improve comet detection and the discovery of non-group comets. In addition to prize money, the winning participants earned discovery credit for the comets they found.
THE POWER OF MANY MINDS FOR DATA SCIENCE INNOVATION
What’s cooler than helping NASA find comets? The different ways members in our Data Science community attacked the problem. Over the past few years, most image processing challenges at Topcoder have been won by deep learning: throwing well-designed neural network architectures at the problem and letting it figure out the solution.
In this challenge, the top two winners didn't use neural networks at all. They applied domain knowledge and easy-to-understand, common-sense rules. There were neural net-based solutions as well among the top seven winners, so these two schools of data science problem-solving both provided comparable results. NASA is thrilled with the results and loves the diversity of perspectives in Topcoder’s community.
For more information about the SOHO Comet Search Challenge, the winners, and Topcoder’s data science capabilities, check out the links below:
SOHO is a cooperative international mission between the European Space Agency (ESA) and NASA’s Goddard Space Flight Center (GSFC). SOHO’s Large Angle and Spectrometric Coronagraph, or LASCO, is an onboard instrument that provides most of the imagery, with two coronagraph telescopes designed to block direct blinding sunlight and observe the much fainter solar corona and solar outflows. As an unintended consequence of LASCO’s sensitivity, LASCO also detects large numbers of previously unknown sungrazing comets. Since it was launched in 1995, the SOHO satellite has detected more than 4,000 new comets. NASA is optimistic that the data-enhancing algorithms developed through the Marathon Match Challenge could have utility beyond comet discovery and tracking for SOHO, and that they may be applicable to coronagraph imagery on other heliophysics observatories.