Department of Computer, Control and Management Engineering

con la presente vi informo che in ottemperanza ai requisiti previsti dalla procedura valutativa per n.1 posto di Ricercatore a tempo determinato tipologia A - SC 09/H1 SSD ING-INF/05 - Dipartimento di Ingegneria Informatica Automatica e Gestionale "A. Ruberti", Codice Bando 2023RTDAPNRR126 , pubblicato su Gazzetta Ufficiale N. 5 del 20.01.2023, si terrà Lunedi' 20 Marzo alle ore 14:00 in aula magna il seminario di Giovanni Trappolini che illustrerà le sue attività di ricerca svolte e in corso di svolgimento. Il seminario sarà anche trasmesso in modalità telematica su Zoom. Per partecipare da remoto connettersi all'indirizzo seguente:

https://uniroma1.zoom.us/j/89155884158?pwd=VngyRUJzanhEc1JoeElycTRtN2QrUT09
Meeting ID: 891 5588 4158
Passcode: 921785

Titolo: Multimodal Neural Databases

Abstract:
The rise in loosely-structured data available through text, images, and other modalities has called for new ways of querying them. Multimedia Information Retrieval has filled this gap and has witnessed exciting progress in recent years. Tasks such as search and retrieval of extensive multimedia archives have undergone massive performance improvements, driven to a large extent by recent developments in multimodal deep learning. However, methods in this field remain limited in the kinds of queries they support and, in particular, their inability to answer database-like queries.
In this talk, we will provide an overview of the consolidated "historical" advances in the field of Neural Databases. We then proceed to explore a new framework, that of Multimodal Neural Databases (MMNDBs). MMNDBs can answer complex database-like queries that involve reasoning over different input modalities, such as text and images, at scale. MMNDBs is the first architecture able to overcome the limitations of both MMIR and vanilla NDB. We compare it with several baselines, showing the limitations of the current state of the art. Preliminary results shown by these techniques show the potential of these new techniques to process unstructured data coming from different modalities, paving the way for future research in the area.

Short bio:
Giovanni Trappolini is a post-doctoral researcher at the Department of Computer Engineering of the Sapienza University of Rome, where he works as a member of the RSTLess Research Group led by Professor Fabrizio Silvestri focusing on  multimodal deep learning. He received his PhD in Machine Learning under the supervision of Emanuele Rodolà, with a thesis on geometric deep learning. In particular, during his doctoral studies, he developed novel algorithms for geometric deep learning, a subfield of machine learning that focuses on learning from non-Euclidean data such as graphs and manifolds. His current research activities focus on applying deep learning techniques to multimodal data, with the aim of developing models that can effectively process and integrate information from different sources such as text, images, and audio. He has published several papers in top-tier conferences and journals, including NeurIPS and ECCV.