DESCARTES - Optimization-Driven Hybrid AI Hybrid Modelling with Effective Domain Adaptation for Robust Prediction (WP3)
Abstract
A CREATE Program on AI-based Decision making in Critical Urban Systems.Coordinators of the work package 3
- Martial Mermillod
- Savitha Ramasamy
- Marina Reyboz
Description
Optimization-Driven Hybrid AI Hybrid Modelling with Effective Domain Adaptation for Robust Prediction.Brief summary : Despite the advancements in deep learning, their robustness to dynamically evolving environments is still a huge challenge. Hence, there is a compelling need to develop deep learning algorithms that are capable of robust representation, especially in real-world data sets that need online adaptation. To this end, we propose developing techniques for domain adaptation, which are capable of detecting and adapting to drifts in data in a self-supervised manner. We borrow principles of learning from lifelong learning realms towards design and theoretical analysis for domain adaptation.
