Hey Sigfrid, looks already pretty good, here a few more comments:
[x] before slide 18, or after slide 3, the Faynman diagram of the neutrino interaction is missing. I would show that, to explain what you mean with EM and HAD shower. And that you initially focussed on hadronic showers which are easier to reconstruct because the LPM doesn’t make thinks more complicated here.
[x] You also might want to show another example at a low SNR to give a better impression of how the signals look like that you’re typically dealing with
[x] To save time, you could remove slide 28 and 29.
The conclusions could also need some sharpening. Here a suggestions: (replace the XX and numbers with the correct values)
[x] •(Deep) Neural network training & optimizing
[x] optimal model similar to VGGXX, XX parameters
[x] better than degree resolution for E>1e18eV
[x] • good results (<3.5deg for E>1e18) also for nu_e-CC interactions
[x] estimated event-by-event uncertainty contours using different noise realizations
[x] There’s lots of potential for improvement and further studies!
[x] Low energy (<1e18eV) performance can likely be improved by more low energy training data
[x] Use DL to quickly estimate performance of different station layouts (-> relevant for Gen2 design) - next step: apply
Hey Sigfrid, looks already pretty good, here a few more comments:
The conclusions could also need some sharpening. Here a suggestions: (replace the XX and numbers with the correct values)