Summer 2018’s been a busy time for the AMVA4NewPhysics network; we’ve had workshops, outreach events, training sessions, meetings, and many more things. I wanted to go through and pick out a few thinks I was involved in.
I guess the summer really kicked off with a workshop-meeting in Athens in an especially warm June. Alongside updates from the researchers, we were delighted to welcome talks from Prof. Gilles Louppe on “Likelihood-free inference, effectively”, and “Reinforced Learning in the Q-learning paradigm” by Mihailo Backovic of B12 consulting.
Whilst the schedule was packed, we luckily still found time to enjoy the historic city, and organise a network dinner with the most fantastic of views.
Heading back to Lisbon I had to prepare for the summer internships my lab organises for university students. As part of the initial training tutorials, I’d been asked to spend an afternoon teaching the students about neural networks. Over the course of three hours I introduced networks and their functionality, and then went through a practical example of how to apply them to a common physics problem.
Two students (Miguel Bengala and Rodrigo Santo) then spent about two months working with my supervisor and I to contribute to performance projections for a forthcoming upgrade to the Large Hadron Collider. Initially they began by continuing to learn about neural networks through implementing their own setup which utilised some recent deep-learning developments to achieve very high scores on data from the 2014 HiggsML challenge. The students learnt very quickly, and were soon able to take concepts from papers and code them up.
Having built up their own frameworks, they then set about the task of optimising classifiers to detect particle collisions in which two Higgs bosons are created; an important process which is expected to be one of the main outcomes of the High Luminosity LHC upgrades. In just two months we were able to complete the study, and document it in an internal note. Alongside similar projection analyses, we’ll be taking it for pre-approval review this Friday!
Part way through the internships I went off to attend the XIIIth Quark Confinement and the Hadron Spectrum conference in Dublin. In a parallel session organised by AMVA4NP’s Tommaso Dorigo and INSIGHTS’ Sergei Gleyzer, called Statistical Methods for Physics Analysis in the XXI Century, I presented some work on applying deep learning techniques to physics data (try it yourself here). Fellow AMVA4NP members Pietro Vischia, Grzegorz Kotkowski, and Pablo De Castro Manzano, also presented their latest research.
Aside from a quick trip to Germany for a friend’s wedding, I mostly had August free to focus on the projection analysis with the students, which was useful since September was extremely busy.
The first event was a day of training in presentation skills in Oxford, by a company called Inside Edge. This involved plenty of practical exercises and presentations, including a filming part where we were able to get detailed feedback, spot mistakes and good things, and really understand where we needed to improve. Overall, it was a great day and I felt I learnt a lot. We also got to meet the newest member of the network, Giovanni Banelli, based in Munich.
After that, I was back in Lisbon for a week with just three days for us to finalise the analysis note before the submission deadline. We made it, luckily! The month then concluded with two back-to-back joint workshops between AMVA4NewPhysics and INSIGHTS.
The first was at CERN, where following an internal review of our outreach activities, we had a day of presentations on the various aspects of scientific outreach. Personally, my main take home realisation is that I understand the purpose and aims of ‘outreach’ when it is termed ‘public engagement’ – ‘outreach’ to me implies making people want to become scientists, but really it seems the purpose is simply to increase people’s interest in, and appreciation of, science, which I think is better described as ‘engagement’.
The day ended with an evening with Tim Blais, AKA Acapella Science, who spoke about the creative process he goes through when taking scientific principles and setting them to the tunes of popular songs.
The second workshop focused on statistical methods for scientific discovery, with presentations not only from some quite renowned scientists and statisticians, but also from younger researchers, who had a dedicated poster session, at which Greg, Pablo, and I presented.
A busy few months indeed, but luckily October is free of travel. At the end of the month we’ll have the last workshop of the network at which all the ESRs will still be present. A date that coincides with the last network deliverable I contribute to. Following that I’ll be off for my industrial secondment with Yandex; three months in Moscow during winter. Brrrrrrr.