Astroinformatics is something of a new concept. Humans have been pointing at the stars since they had fingers. Comparatively, the practice of getting computers to pool through the sky and make discoveries instead is not even half a century old. With the use of machine learning only becoming commonplace in the past decade or so it would be easy to assume that studying it would not feel too high pressure – every technique is new and exciting; every discovery easy to find. This was not to be the case when I began my PhD. Astronomers love coding. It’s one of the big things that attracted me to my project!

When I arrived on the scene machine learning was established practice in astrophysical discovery – sky surveys used them to analyse data; to detect objects, classify them and find where to look next. My current work, which was to improve a model’s ability to detect supernovae in images, seemed to pale in comparison to others. We knew of similar surveys with the same goals as ours, operated by brilliant academics, and it seemed unlikely that I – a man who regularly puts on jumpers the wrong way round – would be able to produce anything of note to the community. It was with this in mind that I tentatively submitted my abstract to the Astroinformatics 2023 conference in Naples for the first week of October.

The LINAS programme afforded me the opportunity to travel to Italy for this conference; one of the first of its kind for the field. Being able to keep a finger on the pulse with regards to the research being done by others and network in person with leaders on the subject is a rare thing for PhD students, and is something that I think the programme encourages extremely well. I was due to present towards the back end of Monday afternoon, which came with its pros and cons. Pro: I had plenty of time to prepare on the day, and I would be able to enjoy the rest of my time in Naples after the fact without worrying about my presentation. Con: I would have to survive a day of listening to experts on Astroinformatics discuss their own work without feeling like I wouldn’t measure up.

As I continued to fret about the upcoming presentation, which would (in my mind) certainly see the end of my career in academia if it went poorly, I tried to focus on reasons why it would go well. I had been given plenty of machine-learning training through LINAS to bring me up to scratch with all the skills required to carry out my work, so in that sense I did have some idea of what I was supposed to talk about. My supervisor and colleagues in the Astrophysics Research Centre had supported my project so far and had agreed that sending me to Italy was a good use of my funds; a promising sign. On top of this, my friends and peers in the Mitchell Institute also seemed interested and engaged in my talk when I explained it to them (or at least were good at pretending to be!). Having such a supportive group of people to discuss your work with is an invaluable thing, and I’m extremely grateful to be part of the friendliest, most encouraging group on campus. It was with the support of all these people that I decided that rather than bolting back down the Capodimonte hill I should probably stick around and give my presentation.

I found that the more I talked about the project, the more I remembered, and the more I realised I actually knew what I was talking about. I stopped trying to memorise what was on my flashcards and started reminding myself of the work I’d done. And when you’re forced to reduce a year of work to a fifteen-minute presentation, you realise how much you’ve done. With the support of an engaged audience it clicked that I didn’t have any need to prove myself; my peers were here to listen to what I had to say because they were genuinely interested. Questions were asked and answered afterwards as easily as casual conversation.

With that, the pressure was off, and I could enjoy the rest of the conference without worrying that I’d be “found out”. I had the amazing opportunity to listen to experts, compare notes with others, and tell my supervisor I was coming back to Belfast with new thoughts and ideas. I still had another four days in Italy though, and a few free afternoons to explore the Amalfi coast…until I’d done that, the next fight with imposter syndrome could wait!

Joshua Weston is in the second year of his LINAS Doctoral Scholarship.  The LINAS Doctoral Training Programme (DTP) seeks to develop a cohort of Doctoral Scholars who can address the implications of massive-scale data processing, artificial intelligence (AI) and machine learning (ML) for both the actual operation of algorithmically driven public decision-making in wider society, and within science and engineering.