Welcome!

I am a third-year PhD student at the Australian National University, focusing on theoretical machine learning and deep learning theory. I am very grateful to be supervised by Stephen Gould, Thalaiyasingam Ajanthan, and Marcus Hutter. Prior to my PhD, I earned a Master’s in Machine Learning and Computer Vision from ANU, where I worked on adversarial machine learning and deep declarative networks under the supervision of Stephen Gould. I also hold a Bachelor of Science (BSc Hons) in Computer Science from Harokopio University of Athens, where I did research on emotion recognition and affective computing using deep learning under Dimitrios Michail and Iraklis Varlamis.

My research lies at the intersection of deep learning, optimisation, and geometry, with an emphasis on building a more rigorous theoretical understanding of neural networks.

Core Research Interests

• The Geometry of Deep Learning: I analyse the geometric structure of the loss landscape and solution space in deep neural networks. My goal is to characterise the properties of optimal solutions in over-parameterised models using tools from differential geometry and frame theory.

• Optimisation and Convergence: I derive convergence guarantees for non-convex optimisation problems in neural networks. My work explores how geometric conditions on the manifold of minimisers can be leveraged to design more efficient and reliable training algorithms.

• Generalisation and Occam’s Razor: My research connects the principle of Occam’s Razor to how and why neural networks generalise. I develop novel bounds by measuring the algorithmic complexity of learned functions, using tools from algorithmic information theory and topology to analyse the structure of their decision boundaries.

I am always open to new collaborations. If you are interested in working together, please feel free to get in touch.

News