معرفی
Carl Allen is a Laplace Junior Chair in Machine Learning at École Normale Supérieure, Paris, working in the research group of Stéphane Mallat, Giulio Biroli and Garbiele Peyré. Previously, he was a postdoctoral fellow at ETH Zurich and completed his PhD in Machine Learning in 2021 at the University of Edinburgh under the supervision of Professors Tim Hospedales and Iain Murray.
His educational background includes a BSc in Mathematics & Chemistry from the University of Southampton, an MSc in Mathematics and the Foundations of Computer Science (MFoCS) from the University of Oxford, and MScs in Artificial Intelligence and Data Science from the University of Edinburgh. Before transitioning to AI/ML research, he spent several years in Project Finance.
Allen's research focuses on mathematically understanding mechanisms behind successful machine learning methods, particularly neural networks. He investigates how machine learning models exploit aspects of data distribution from a probabilistic perspective. His current topics include explaining how VAEs disentangle independent factors of data, identifying mathematical models behind self-supervised learning, and deriving probabilistic interpretations of softmax classification. His PhD work investigated neural representations of discrete objects and their relationships, with a main result explaining how word embeddings can seemingly be added and subtracted (e.g., queen ≈ king - man + woman), which received Best Paper (honorable mention) at ICML 2019.
His research spans theoretical foundations of machine learning, with particular emphasis on representation learning, disentanglement, and probabilistic modeling of neural networks. His work connects mathematical principles with practical machine learning applications, aiming to develop more interpretable and reliable algorithms.
Allen has received notable recognition including a Best Paper honorable mention at ICML 2019 and a research grant from the Hasler Foundation. He has delivered invited talks at prestigious institutions including Harvard Center of Mathematical Sciences & Applications and Astra-Zeneca.
His collaborative work spans multiple institutions, including a notable internship at Samsung AI Centre, Cambridge, where he worked at the intersection of representation learning and logical reasoning. His research has significant implications for developing more interpretable, reliable, and theoretically grounded machine learning systems.
