I am an Assistant Professor (lecturer) in Machine Learning at the Department of Mathematics at Imperial College London, affiliated with Imperial-X.
Very broadly speaking, I am interested in the principles that underlie cognition, both in nature and artificial. Some specific themes we work on:
- Computationally constrained learning. A recent and popular approach in the ML community is to simply increase the size of models and datasets (GPT and the like). In this case, the computational cost of running and training the models becomes a bottleneck. We look at: few-shot learning, spiking neural networks, optimise subsets of parameters or use only parts of the dataset (coresets), biologically plausible learning algorithms.
- Training procedures for more robust learning. The other approach is to cleverly modify the training procedures to learning more robust and generalisable representations of data. We look at: modifying the way we represent the data, contrastive loss approaches, active inference and Bayesian methods to represent uncertainties.
We work in close collaboration with neuroscientists to inspire new methodologies, as well as to apply ML methods to drive understanding into the mechanisms of the brain.
See Google Scholar for a full overview of publications or see the Research page for recent publications.
- Email: firstname.lastname@example.org
For PhD students and postdoctoral researchers, send me an email with the subject line ‘building better ML’ if you are interested or directly apply through Imperial’s website.
- Google Scholar
This is a jekyll based resume template made by Ankit Sultana