Vicente Castro Solar

I am writing my Master’s Thesis under the supervision of Manuel Roveri on using Neural Architecture Search (NAS) for Continual Learning problems. Broadly, I am studying plasticity and stability in deep neural networks to find the best model expansion to be able to learn new samples, but also retain the previously learnt data.

I have two other research projects going on with the iHealth Institute:

• Studying how Meta-Learned Initialisations affect the performance of Implicit Neural Representations for MRI imaging. Through our method we reduced the training time by x6 factor. These and some other interesting results will be available in the incoming paper!

• In a more recent project, we are proposing an extension of Gaussian Splatting for the K-Space (or Fourier Space). This allows us to perform a joint optimisation over both domains, improving performance and training times over previous method.