My first main PhD project is using a regularised conditional generative adversarial network (rcGAN) and applying it to convergence map reconstruction. Recent developments in cGANs mean the model is easy to train, and suffers no mode collapse. This code is for the planar setting, and has been tested on the COSMOS field data. A paper will be out on this very shortly, and the code will be publicly available.
My next project will focus on spherical mass mapping, watch this space for more details.
S2WAV is a Python package for automatically differentiable wavelet transforms on the sphere. The package is headed by Matt Price, and it was great experience to help develop this package alongside other collaborators Alicja Polanska, and Prof. Jason McEwen.
View the code View the paperI presented my work on mass mapping with regularised conditional GANs at COSMO21 in May, 2024. You can find my slides below.
View my talkSo grateful to get the chance to present a poster at the cosmology session at Moriond 2024. Such a stacked conference, with brilliant talks every day. Not to mention, great skiing buddies! The conference had an amazing atmosphere, and I hope I can return in the future to present more of my work!
View my poster View the conference proceedingsAIUK was an eye-opening conference I attended in March, 2024. Given how quickly AI is becoming part of daily public life, I learned a lot about AI policy, ethics, and what the future of AI may hold. A brilliant conference, with so many influential speakers - thank you to the Alan Turing Institute for hosting such a great event.
I presented a poster as well as a flash talk at the Machine Learning session at EAS 2023, talking about initial preliminary results using conditional GANs for mass mapping. I'm grateful I had the opportunity to attend such a large conference so early into my Ph.D., and found the talks absolutely fascinating!