Research

My main research line revolves around the role of data structure in Hopfield networks. Since we found a series of promising phase transitions (learning, generalization), my plan for the near future is to look for the same behaviour in modern models in deep learning. The most promising candidates are the attention mechanism and generative diffusion, that have been recently connected to modern Hopfield networks (also known as dense associative memories).

During my previous post-doc and my PhD I studied topics that still interest me, especially because of their wide scope and their multidisciplinarity.