Hi! I'm a fourth-year PhD student in machine learning & computer vision at Queen Mary University of London (QMUL), working with Ioannis Patras, Mihalis A. Nicolaou, and Yannis Panagakis. I'm also an Honorary Associate (visiting scholar) at the University of Wisconsin–Madison, working with Grigorios Chrysos. Previously, I was fortunate enough to intern at both Huawei Noah's Ark Lab London and The Cyprus Institute.

My PhD research focuses on interpretable and controllable machine learning. I'm interested in both discovering interpretable mechanisms in existing models and designing new layers/architectures that better exhibit such properties from the outset. A central theme of my work is leveraging tensor methods and structured matrices to break down machine learning models’ computations into human-understandable subcomponents. My ultimate goal is to advance AI safety and trustworthiness by making machine learning systems more transparent.

News

• [2024.11] Recognized as a Top Reviewer at NeurIPS'24
• [2024.09] 1 paper accepted at NeurIPS'24 (on scalable expert specialization in MoEs)
• [2024.09] Started at UW-Madison as an Honorary Associate
• [2024.06] 1 paper accepted at TPAMI (extending our ICLR'23 work)
• [2023.09] 1 paper accepted at NeurIPS'23 (on semantic subspaces in vision-language models)
• [2023.07] Started as a PhD Research Intern with Jiankang Deng at Huawei R&D UK
• [2023.01] 1 paper accepted at ICLR'23 (on parts/appearance decomposition in GANs)
• [2021.09] Started as a PhD student at QMUL

Selected publications

• "Multilinear Mixture of Experts: Scalable Expert Specialization through Factorization"
  J. Oldfield, M. Georgopoulos, G. G. Chrysos, C. Tzelepis, Y. Panagakis, M. A. Nicolaou, J. Deng, I. Patras
  NeurIPS, 2024
  [pdf | code | project page]
• "Parts of Speech–Grounded Subspaces in Vision-Language Models"
  J. Oldfield, C. Tzelepis, Y. Panagakis, M. A. Nicolaou, I. Patras
  NeurIPS, 2023
  [pdf | code | project page]
• "PandA: Unsupervised Learning of Parts and Appearances in the Feature Maps of GANs"
  J. Oldfield, C. Tzelepis, Y. Panagakis, M. A. Nicolaou, I. Patras
  ICLR, 2023 (& TPAMI 2024)
  [pdf | code | project page]
• "Tensor methods in computer vision and deep learning"
  Y. Panagakis, J. Kossaifi, G. G. Chrysos, J. Oldfield, M. A. Nicolaou, A. Anandkumar, S. Zafeiriou
  Proceedings of the IEEE, 2021
  [pdf | code]

Experience

• [2024.09–2024.12] Honorary Associate (University of Wisconsin–Madison, Madison)
• [2023.07–2024.01] Research Intern (Huawei Noah's Ark Lab, London)
• [2019.11–2020.09] Research Intern (The Cyprus Institute, Nicosia)

Invited talks

• [2024.06] Tensor Decompositions in Large Scale Deep Learning (Archimedes Research Unit, Athens)

Teaching

• [2024–2024] Teaching assistant for Deep Learning and Computer Vision (QMUL, ECS795P)
• [2021–2022] Teaching assistant for Machine Learning (QMUL, ECS708)
• [2021–2021] Teaching assistant for Artificial Intelligence (QMUL, ECS629)

Awards

• [2024] Top Reviewer: NeurIPS 2024
• [2021] Full EPSRC Studentship (QMUL): 3.5 years' PhD funding
• [2018] Academic Excellence (£50; HEAR): Ranked #1 in CS department for undergrad