About

I am a DPhil candidate at the University of Oxford and at the Alan Turing Institute.

I am interested in two main strands of research.

The first focuses on studying deep probabilistic models, specifically their adversarial robustness and their ability to build representations that are factorisable and human-interpretable.
The second uses functional analysis and Langevin dynamics to study commonly used regularisation methods, such as Gaussian noise injections.

Fractal Structure and Generalization Properties of StochasticOptimization Algorithms, A Camuto, G Deligiannidis, M Erdogdu, M Gürbüzbalaban, U Şimşekli, L Zhu - arxiv, 2021
Variational Autoencoders: A Harmonic Perspective, A Camuto, M Willetts - arxiv, 2021
Certifiably Robust Variational Autoencoders, B Barrett, A Camuto, M Willetts, T Rainforth - arxiv, 2021
Asymmetric Heavy Tails and Implicit Bias in Gaussian Noise Injections, A Camuto’, X Wang’, L Zhu, C Holmes, M Gürbüzbalaban, U Şimşekli - International Conference on Machine Learning, 2021
Learning Bijective Feature Maps for Linear ICA, A Camuto’, M Willetts’, B Paige, C Holmes, S Roberts - International Conference on Artificial Intelligence and Statistics, 2021
Towards a Theoretical Understanding of the Robustness of Variational Autoencoders, A Camuto, M Willetts, S Roberts, C Holmes, T Rainforth - International Conference on Artificial Intelligence and Statistics, 2021
Improving VAEs’ Robustness to Adversarial Attack, M Willetts’, A Camuto’, T Rainforth, S Roberts, C Holmes - International Conference on Learning Representations, 2021
Explicit Regularisation in Gaussian Noise Injections , A Camuto, M Willetts, U Şimşekli, S Roberts, C Holmes - Advances in Neural Information Processing Systems, 2020

^{’ equal contributions}

Winner Web3 Weekend 2021
Winner ETHOnline 2020
Winner HackFS 2020
Winner Imperial College’s Stephen Richardson Award 2017 - for the best MEng thesis
Winner Imperial College’s Governor Prize 2017 - for the best graduating student

In my spare time I take pictures.