We often use stochastic descriptions to model complex systems. Many of our projects involve kinetic Monte Carlo, agent-based models and multiscale modeling. A reoccurring theme is how to leverage distributed e-infrastructure for simulations and how to use machine learning to construct approximations.
A core theme in the group is the use of machine learning to make scientific computing software and infrastructure more efficient, interactive and scalable. We also do disciplinary research in specific areas of ML, such as likelihood-free inference and privacy-preserving federated machine learning.
Our research in distributed computing and data engineering sciences ranges from development of new ways to manage large and fast data to design and development of massively parallel, interactive, cloud native applications operating in cloud, fog and edge infrastructure.
The group participates in the eSSENCE strategic initiative on eScience, focusing on eScience tools and technologies.
Scaleout Systems is a spin-off from the group, focusing on privacy-preserving AI and cloud-native machine learning.
New version of FEDn, our open framework for scalable federated learning, is out! #federatedlearning
This release introduces new functionality such as a new events view, new network layout, (reducer, combiner and clients hierarchy), a more straight forward setup guide, and a lot more. Take a look for your self!
https://github.com/scaleoutsystems/fedn/releases/tag/v0.2.4
Here's an opportunity to learn about federated learning in an hands-on workshop. Sign up below!
https://www.scilifelab.se/event/fedml/
#ml #ai #machinelearning https://www.scilifelab.se/event/fedml/
