Data Science for Industry and Physics
A research unit of the Digital Industry center at FBK, on the forefront of data science and artificial intelligence
Who we are
Our multifaceted team of experts develop groundbreaking solutions for analyzing and predicting time series and spatiotemporal data, elevating precision and dependability across diverse fields. By leveraging the capabilities of deep learning, we aim to provide a versatile toolkit that spans various disciplines to effectively tackle the unique challenges presented by each domain.
Active Areas Of Research
DSIP is actively collaborating with research organizations, public institutes and industrial partners across a wide variety of fields
Industrial Sector
Integrative AI
We are dedicated to supporting industries in their digital transformation through optimizations and data-driven insights. Our focus lies at the forefront of Integrative AI, where we seamlessly blend the strengths of deep learning and formal methods to enhance the performance of solutions tailored for industrial projects.
Predictive Maintenance and Condition Monitoring
By harnessing the adaptability of deep learning alongside the representation of background knowledge using formal methods, we aim to create a symbiotic synergy that optimizes model performance across various applications. This amalgamation enables precise insights and facilitates real-time data analysis of equipment health.
Digital Twins
Our research extends to developing digital twins, where the fusion of deep learning and formal methods enhances the fidelity and reliability of virtual representations of physical systems. Through these interdisciplinary approaches, our ultimate goal is to redefine the boundaries of what is achievable in industrial AI, setting new standards for efficiency, reliability, and innovation.
Weather & Climate
AI-based precipitation nowcasting
We specialize in creating artificial intelligence techniques for precipitation nowcasting, which refers to the short-term precipitation forecast (0-6 hours). Our models leverage meteorological radar data to enhance the accuracy of precipitation forecasts in terms of spatial and temporal resolution, as well as temporal extension.
Early warning systems
We specialize in creating early-warning systems for populations and groups exposed to environmental risks. Our systems aggregate and process data from multiple sources in real time. We aim to develop a fully automated intelligent alert systems. These systems provide personalized alerts to individual users through various distribution channels, including web apps, chatbots, email, and SMS texts.
AI for Earth and Climate
Within the Spoke 4 of the National Centre for HPC, Big Data and Quantum Computing, DSIP develops state-of-the art solutions to improve the results from the Earth System Models, generating both weather forecasts and climate projections. Two applications are currently under developement: the DL-based downscaling of weather forecasts and climate scenarios; and the enhancement of Earth System models parameterizations with ML-based models.
Physics Research
Pruning Deep Learning networks
We have made significant strides in addressing the challenge of optimizing deep neural networks. Our team has successfully devised an algorithm introducing a groundbreaking automated pruning approach. This innovative algorithm is designed to automatically prune a specified fraction of filters and nodes within a deep neural network while vigilantly managing performance loss. It contributes to a more computationally efficient model and aligns with the broader goal of creating leaner and more resource-efficient deep neural networks. Developed in activities connected with the ATLAS experiment at CERN, the methods find application transversally to all our research fields.
Space weather
We aim to use deep learning to look into potential earthquake precursors in connection with the LIMADOU contribution to the CSES mission. Our focus centers on examining variations in elements of the ionosphere as potential indicators of seismic activity. Another line of research is connected with DL data modeling to predict DST fluctuations in an attempt to contribute to developing an early warning system for geomagnetic storms.
Reach out
Email us
marco.cristoforetti@fbk.eu
Call us
+39 0461 314 553