- Docente responsabile
- MARCO DOMENICO SANTAMBROGIO
- CCS proponenti
- Ingegneria Informatica
- CFU
- 2
- Ore in presenza
- 10
- N° max studenti
- 100
- Parole chiave:
- Federated Learning, Machine Learning, Time Series Analysis from Wearables, wellbeing, wellness
- Tag
- Computer science, Engineering, Sport, Health and lifescience, Information technologies
Descrizione dell'iniziativa
The growing availability of heterogeneous data is radically transforming the way we can observe and interpret individual well-being. Access to this data is now possible both through commercial wearable devices, thanks to APIs and SDKs that enable interoperability, and through the development of customized solutions based on low-cost microprocessors and platforms capable of acquiring and processing signals from environmental and physiological sensors.This lays the foundation for new forms of personalized wellness monitoring in homes, workplaces, and shared spaces. Among the most promising applications are sleep analysis, cognitive performance, and sports performance. However, the integration and analysis of these multimodal time series pose new computational challenges, especially when data is generated on devices with limited resources. It is therefore necessary to develop solutions that minimize dependence on the cloud while preserving user privacy.To address these challenges, it was necessary to develop solutions capable of performing inference and machine learning directly on the device, reducing the central computational load and the exchange of sensitive data. This is where the Federated Learning paradigm comes in, allowing distributed models to be trained while keeping data locally on edge nodes. This enables the development of intelligent architectures capable of responding dynamically to context and individual needs.The course offers a comprehensive overview of the process: from the collection and pre-processing of multimodal data, including the integration of information gathered from commercial wearables (e.g., Garmin, Polar, Apple Watch), exploiting the dedicated APIs and SDKs, to the design of lightweight models for on-device inference, and finally to the implementation of federated strategies for distributed training.
Periodo di svolgimento
dal November 2025 a December 2025
Calendario
Il calendario del corso è visibile sul calendario: tinyurl.com/PiAatDEIB
Per le lezioni di questo corso si cerchi #BALANCE
- 13/11/2025 - 12:30/14:00
- 20/11/2025 - 12:30/14:00
- 24/11/2025 - 12:30/14:00
- 01/12/2025 - 12:30/14:00
- 04/12/2025 - 12:30/14:0
Le aule saranno le stesse del corso di CSI (Creativity, Science and Innovation).
Potrete trovarle a questo URL: https://santambrogio.faculty.polimi.it/dida/csi/index.htm
O a questo calendario: https://santambrogio.faculty.polimi.it/dida/csi/calendario.htm