Emanuele Rodolà
Full Professor of Computer Science, Sapienza University of Rome
Prof. Emanuele Rodolà is a Full Professor of Computer Science at Sapienza University of Rome, where he leads the GLADIA group focusing on Geometry, Learning, Audio, and Applied AI. His work in this area is funded by an ERC Starting Grant and a Google Research Award.
Before his current position, he served as an Assistant and then Associate Professor at Sapienza (2017-2020), a postdoc at USI Lugano (2016-2017), an Alexander von Humboldt Fellow at TU Munich (2013-2016), and a JSPS Research Fellow at The University of Tokyo (2013), in addition to visiting periods at Tel Aviv University, Technion, Ecole Polytechnique, and Stanford.
He is a fellow of ELLIS and the Young Academy of Europe. Prof. Rodolà has received several awards for his research, he has been active in the academic community, serving on program committees and as an area chair for leading conferences in computer vision, machine learning, and graphics (CVPR, ICCV, ICLR, NeurIPS, etc.). His current research mainly focuses on neural model merging, representation learning, ML for audio, and deep multimodal learning, and has published around 150 papers in these areas.
Lecture
Unlocking Neural Composition
Ideally, the distribution of the latent representations within any neural network should depend only on the task, the data, the loss, and other architecture-specific constraints. However, factors such as the random weights initialization, hyperparameters, or other sources of randomness may induce incoherent latent spaces that hinder any form of reuse. In this talk I’ll report an important (and somewhat surprising) empirical observation: under the same data and modeling choices, distinct latent spaces typically differ by an unknown quasi-isometric transformation; that is, in each space, the distances between the encodings do not change. I’ll then show how simply adopting pairwise similarities as an alternative data representation leads to guaranteed isometry invariance of the latent spaces, effectively enabling latent space communication: from zero-shot model stitching to latent space comparison between diverse settings. Several validation experiments will follow on different datasets, spanning various modalities (images, text, graphs), tasks (e.g., classification, reconstruction) and architectures (e.g., CNNs, GCNs, transformers).
