Energy-efficient algorithms and Machine learning

We develop novel algorithms for resource-constrained intelligent devices. This involves optimizations at different levels, including:

  1. data acquisition, by using techniques such as compressed sensing or event-based sampling;
  2. algorithm execution, by exploiting modern edge/fog/cloud architectures to distribute machine-learning methods in a robust and efficient way; and
  3. data transmission, by processing the data as close as possible to where it is produced and reducing communications to high-level features and results.

Automatic Optimization Flow for Facebook’s Deep Learning Recommendation Model


Edge AI-deployed DIGItal Twins for PREDICTing disease beyond COVID-19

Embedded AI for aerospatial navigation

Energy-efficient acquisition and embedded processing of bio-signals

Federated machine learning over fog/edge/cloud architectures

HDTorch – a PyTorch-based library for HyperDimensional Computing

Multimodal and personalised methods for health and wellness monitoring