Overview
The large quantity of data available on the internet has led to the deployment of extremely powerful artificially intelligent models. Such models, however, require an impressive amount of computational resources for training and inference. This limits the deployment of such models in a real-world scenario outside the sandbox of massive data centers.
Inspired by the capability of human brains to learn new concepts efficiently, the main activity of the group is to research novel training techniques that reduce the computational burden of neural models, improving the ability to learn new concepts continuously and extract knowledge from different modalities.
Research directions
The key topics of interest are:
- Low supervision models
- Self-supervision
- Cross-modal learning
- Incremental Learning
- Debiasing neural networks
- Domain Adaptation and Generalization
Members
- Paolo Rota, Principal Investigator
Publications
For a complete list see Paolo Rota scholar profile.