NUBIS is participating in CCGRID 2025, presenting its contributions to the MLSysOps initiative (vAccel) at the RISE workshop. The team is showcasing their latest paper, titled: “MLIoT: Transparent and Secure ML Offloading in the Cloud-Edge-IoT Continuum”
Excited to share that our paper, “TMModel: Modeling Texture Memory and Mobile GPU Performance to Accelerate DNN Computations,” will be presented by Jiexiong Guan (UTH, University of William & Mary) at the International Conference of Supercomputing, which will take place in Salt Lake City, U.S.A, on June 8-11, 2025. This work is a collaboration between UTH, the University of William & Mary, and the University of Georgia.
Mainstream mobile GPUs (such as Qualcomm’s Adreno) usually have a 2.5D L1 texture cache that offers throughput superior to that of on-chip memory. However, to date, there is limited understanding of the performance features of such a 2.5D cache, which limits their optimization potential. TMModel introduces a novel performance modeling framework for mobile GPUs that combines micro-benchmarking, an analytical performance model, and a lightweight compiler to optimize DNN execution based on access patterns and GPU parameters. TMModel delivers up to 66× speedup for end-to-end on-device DNN training with significantly lower tuning cost than existing frameworks. As mobile devices grow more powerful, this work is a step towards efficient, real-time deep learning training directly on such devices.
Two highly productive days in Athens for the 5th MLSysOps plenary! On March 20–21, the consortium came together to sync up on technical progress and engage in hands-on discussions around system integration, machine learning, and open-source development.
Top highlights from the plenary:
On January 22, the Machine Learning for Edge Computing Systems (ML4ECS) Workshop took place as part of HiPEAC. The workshop provided a platform for leading researchers, industry experts, and practitioners to discuss the latest advancements and challenges at the intersection of machine learning and edge computing systems.
Throughout the day, we had the privilege of hearing interesting presentations, engaging in interactive discussions, and exchanging ideas. We would like to thank all speakers, participants, and co-organizers from the CODECO and EDGELESS projects who contributed to the workshop’s success. We look forward to building on the momentum and fostering collaboration in this rapidly evolving field.
The Final Program for the ML4ECS Workshop is online!
Join us as we dive into the cutting-edge intersection of Machine Learning and Cloud-Edge computing.
What to Expect:
✅ Keynote from leading expert in ML and Cloud-Edge systems
✅ Interesting research presentations and Demos
✅ Engaging discussions on the future of ML-driven innovation.
📅 When: January 22, 2025
📍 Where: Palau de Congressos @ Fira de Barcelona as part of the HiPEAC conference
Don’t miss the opportunity to connect with researchers, innovators, and industry leaders shaping the future.
🔗 Check out the program and join us: https://ml4ecs.e-ce.uth.gr/
Nikos Bellas (UTH) presented MLSysOps at the 1st EDGELESS Info Day: Challenges and Opportunities in Cognitive Edge Computing.
A recording of the event is available on the EDGELESS youtube channel:
Ildi Alla et. al. of INRIA will be presenting a paper titled “Robust Device Authentication in Multi-Node Networks: ML-Assisted Hybrid PLA Exploiting Hardware Impairments” at the Annual Computer Security Applications Conference (ACSAC) which will take place in December 2024 at Waikiki, HI, USA.
Ubiwhere was again this year at the Portugal Smart Cities Summit which took place in Lisbon from 8-10 October and at the Smart City Expo World Congress in Barcelona on November 5-7. MLSysOps was presented as part of the research projects Ubiwhere is involved in that are working towards improving quality of life in Cities.
On Oct 24 and 25, Ubiwhere’s headquarters welcomed the members of the MLSysOps Project consortium for the 4th plenary meeting. 2 days of sharing information about the project, its technological advances and next steps.
On October the 23rd, NTT DATA Italia participated in the Demo session at the Automotive Edge Computing Consortium (AECC) meeting in Berlin, hosted by Toyota Motor Corporation, Toyota Motor Europe, and GSMA. NTT DATA Italia presented the vision for leveraging the GSMA Open Gateway to enhance vehicle mobility, focusing particularly on roaming and future applications of Federation among Communication Service Providers (CSPs) and Telematics Service Providers (TSPs). This approach facilitates seamless access to the best-matching Edge/MEC resources to meet the specific application needs of automotive OEMs.
The solution also emphasized the role of AI and ML technologies, which NTT DATA Italia is actively exploring within the MLSysOps Project as part of the EU Horizon projects. This innovative approach generated strong interest from external participants, including members of the 5G Automotive Association (5GAA) attending the AECC meeting.
Skip to content