I wrote this paper with NASA and Princeton on predicting magnetic flux during solar active region emergence.
problem statement: is it possible to predict the magnetic flux curve before the emergence of the active region itself becomes visible?
this matters because active regions drive space weather events, and earlier warnings can help forecasting systems react before the surface signal is obvious.
What seemed particularly intriguing about the work was the fact that the simpler model outperformed the more complex one.
the stacked LSTM with hybrid loss outperformed the encoder-decoder architecture due to the fact that the task at hand was not only predicting the correct value of flux, but also its timing.
the model predicts the magnetic flux emergence 3-10 hours before observation time based on acoustic power and continuum intensity from SDO/HMI data.
https://t.co/WJPRdJqlIK
Just cold approached a guy in my boarding line. On my way to my internship.
Turns out they do smth similar to what I'm gonna be working on. The best part: they hire interns.
Don't be scared!
Now I gotta ask the pilots for a pic in the cockpit.