Prepare presentation on monday

[Kladdy]

"I also checked with Anna and we will have a reconstruction focussed call next Monday 7pm. It would be great if you could present your results. It would be good if you could send me your slides in advance, then I can give you feedback etc. Your ARIANNA talk is a good start. I think you only need to update it a bit with the newer results

yes. but it is not strict, could also be 20min. So you don’t need to train giving the talk to get to exactly 15min."

[Kladdy]

• [x] remove slide 25. Just say that you optimized the architecture for a long time.

• [x] slide 26: First show results just for one data set (Alvarez2009 hadronic). On half the slide, characterize the data set (i.e. which Askaryan model, what station configuration = a shallow station with 4 LPDAs + 1 dipole, show nu-energy histogram). Then show the on the right side the obtained resolution for all events as a histogram. Then say that this resolution is difficult to interpret because the energy distribution is not representative, so you analyzed the resolution in more detail

• [x] Then show on the next slide the performance dependence on different paramters just for this dataset. So a combination of slides 30-33.

• [x] You can also speculate on this slide (I mean you can say) that the low energies perform worse because they are underrepresented in the data set -> we’ll fix that in the future to improve resolution at low E.

• [x] Also on the earlier slide 26 say that you did not do any quality cuts. You used all triggered events

• [x] Then you can move on with slide 28 (the dependence on the number of training data)

• [x] Then move on with the different data sets (ARZ2020). This is really an exiting result that you get almost the same resolution with the more complex emission model (so the current slides 26.

• [x] You can also show slide 27 as a cross check. But probably not that important.

• [x] Slide 35: needs more introduction. Mention that you now reconstruct the same event with 10,000 different noise realizations to obtain the uncertainty on a single event level. Maybe you can also add another zoom-in plot. Do you have another example with an even better angular resolution?

• [x] remove the intermediate title slides (e.g. 34, 36

• [x] slide 41: remove “report writing” this is not relevant for this presentation. Rather try to give an executive summary of the results. Try it yourself, this will be a good training for you. As next steps I would say that we plan to apply the same to the deep station layout to check its reconstruction performance.

• [ ] You will also need to generate a pdf of the talk that we can upload

• [ ] please send me an updated version when you get to it

Hey Sigfrid, the presentation looks already really good. It will be a great talk. Above I gave you a list of relatively detailed suggestions to further improve the talk, so that it is better matched to the background of most people. E.g. so far we only used the Alvarez2009 model with hadronic showers. And other people (e.g. Ilse) who will also present this Monday finds that with the more complex model ARZ2020 everything becomes much more difficult. So by showing at the end of the talk that it works equally well for the ARZ model you give an exiting message

we can also talk (zoom) if you have more questions