This year's SETI class has gone through the curriculum faster than any previous class. After five weeks of instruction and exercises, 5 out of our 6 student teams have already detected signs of intelligence in a massive data set. The detection of the Voyager 1 signal is a stepping stone in the course because it requires a good grasp of the theoretical material, including Fourier transforms, Doppler shift, and noise statistics, as well as the development of sufficient programming skills. I attribute the accelerated pace in large part to a suggestion by Robert Geil, a SETI course alumnus who recently obtained his bachelor's degree in computer science and who continues to work in the UCLA SETI Group. Robert proposed "pair programming," a concept that he learned during one of his internships. In this mode, the 3 or 4 students in each team rotate which team member shares his or her screen and writes code, while the other team members contribute suggestions in real time. We have noticed that this scheme results in more vibrant interactions between the students and a more rapid convergence to the correct solutions. A rising tide lifts all boats. Other contributors to the accelerated pace may be a good distribution of Python expertise among teams and course materials that improve every year. We are delighted by the rapid progress because the students will have more time to focus on data analysis and code development.
When he is not assisting with the course, Robert is also improving our data processing pipeline. For instance, he wrote a graphical processing unit (GPU) version of the algorithm that sums our power spectra in each of 200 million channels with more than 1000 trial values of frequency drift rate. Robert has decreased the computing time for this task from 2 days to under 3.75 hours per hour of telescope time. This improvement brings us dramatically closer to a real-time data processing pipeline. With a slightly more powerful GPU and/or a second workstation, we should be able to execute this task faster than it takes to collect the data.
In my next newsletter, I will tell you about our most recent observing run with the Green Bank Telescope, which was a great success last week. We downloaded the first 4 terabytes of our data in about 60 hours.
Finally, I am excited to share some results that appeared in the May 2021 issue of Nature Astronomy. We used radar observations spanning 15 years to reveal new insights about the interior and atmosphere of our sister planet, Venus. UCLA wrote an informative press release about our work, and the full article is available on arXiv (view and download) and Nature Astronomy (view only).
Warm regards,
Jean-Luc Margot
|
|
