Application Aware, Life-Cycle Oriented Model-Hardware Co-Design Framework for Sustainable, Energy Efficient ML Systems

Informations projet

SustainML

N° de convention de subvention: 101070408

DOI

10.3030/101070408

Date de signature de la CE 20 Juin 2022

Date de début 1 Octobre 2022

Date de fin 30 Juin 2026

Financé au titre de

• Digital, Industry and Space

Coût total

€ 3 742 855,00

Contribution de l’UE

€ 3 742 855,00

Coordonné par

PROYECTOS Y SISTEMAS DE MANTENIMIENTO SL
Spain

CORDIS fournit des liens vers les livrables publics et les publications des projets HORIZON.

Les liens vers les livrables et les publications des projets du 7e PC, ainsi que les liens vers certains types de résultats spécifiques tels que les jeux de données et les logiciels, sont récupérés dynamiquement sur OpenAIRE .

Livrables

Documents, reports (8)

Explainable AI for decision making - Initial

A set of working examples of more context-dependent explainable AIs that allow intelligent systems to express trade-offs relevant to the current task more effectively than current approaches. An online tutorial that explains our method to enable other AI researchers to use this type of human-computer collaboration approach.

Report on communication, exploitation and dissemination - M18

Report on communication, exploitation and dissemination

Hardware accelerator architecture report

Report on the detailed analysis of suitable cross-layer optimizations used to design low-power and energy-efficient hardware accelerator architectures.

Annual report on collaboration - Y1

Annual report on collaboration

Report on communication, exploitation and dissemination - Initial

Report on communication, exploitation and dissemination

Task modeling from user definition results - Initial

Report on ML task stage and parameter space exploration, node graph descriptions, and results in problem modeling from user descriptions.

Framework architecture, back-end, front-end design and release - Y1

Reports on the frame architecture design that integrates all software interfaces from WP1-WP4, and software release of the back-end and front-end of the framework.

Exploring new ML models Report

Report on new interactive visualization methods to assist experienced and inexperienced ML developers in better understanding the design space of available ML models and in particular their relationships, benefits, and resource costs.

Demonstrators, pilots, prototypes (1)

Carbon footprint based model optimization tool

Data Management Plan (2)

Data Management Plan - M18

Data Management Plan

Data Management Plan - Initial

Data Management Plan

Publications

Other (1)

Adaptive Conformal Regression with Jackknife+ Rescaled Scores

Auteurs: Deutschmann, Nicolas; Rigotti, Mattia; Martinez, Maria Rodriguez
Publié dans: 2023, ISSN 0000-0000
Éditeur: Transactions on Machine Learning Research, TMLR 2024
DOI: 10.48550/arxiv.2305.19901

Peer reviewed articles (7)

Mixed selectivity: Cellular computations for complexity

Auteurs: KM Tye, EK Miller, FH Taschbach, MK Benna, M Rigotti, S Fusi
Publié dans: NEURON, 2024, ISSN 1097-4199
Éditeur: NEURON
DOI: 10.1016/j.neuron.2024.04.017

Ten recommendations for reducing the carbon footprint of research computing in human neuroimaging

Auteurs: Nicholas E. Souter, Loïc Lannelongue, Gabrielle Samuel, Chris Racey, Lincoln J. Colling, Nikhil Bhagwat, Raghavendra Selvan, Charlotte L. Rae
Publié dans: Imaging Neuroscience, 2023, ISSN 0270-6474
Éditeur: MIT Press Direct
DOI: 10.1162/imag_a_00043

Efficient Self-Supervision using Patch-based Contrastive Learning for Histopathology Image Segmentation

Auteurs: Boserup, Nicklas; Selvan, Raghavendra
Publié dans: Boserup , N & Selvan , R 2023 , Efficient Self-Supervision using Patch-based Contrastive Learning for Histopathology Image Segmentation . in Proceedings of the Northern Lights Deep Learning Workshop 2023 . Septentrio Academic Publishing , Proceedings of the Northern Lights Deep Learning Workshop , vol. 4 , 2023 Northern Lights Deep Learning Workshop - NLD 2023 , Tromsø , Norway , 10/01/2023 . https://doi.org/10.7557/18.6798, 2023, ISSN 2703-6928
Éditeur: Proceedings of the Northern Lights Deep Learning Workshop
DOI: 10.48550/arxiv.2208.10779

A Knowledge Distillation Framework for Multi-Organ Segmentation of Medaka Fish in Tomographic Image

Auteurs: Bhatt, Jwalin; Zharov, Yaroslav; Suh, Sungho; Baumbach, Tilo; Heuveline, Vincent; Lukowicz, Paul
Publié dans: Crossref, 2023, ISSN 1945-8452
Éditeur: IEEE
DOI: 10.48550/arxiv.2302.12562

Energy-Efficient, Low-Latency and Non-contact Eye Blink Detection with Capacitive Sensing

Auteurs: Mengxi Liu, Sizhen Bian, Zimin Zhao, Bo Zhou, Paul Lukowicz
Publié dans: Frontiers in Computer Science, 2024, ISSN 1664-3224
Éditeur: Frontiers
DOI: 10.3389/fcomp.2024.1394397

Operating Critical Machine Learning Models in Resource Constrained Regimes

Auteurs: Selvan, Raghavendra; Schön, Julian; Dam, Erik B
Publié dans: Medical Image Computing and Computer Assisted Intervention – MICCAI 2023 Workshops ISBN: 9783031474248, 2023, ISSN 1572-8080
Éditeur: Springer
DOI: 10.48550/arxiv.2303.10181

Novel adaptive quantization methodology for 8-bit floating-point DNN training.

Auteurs: Hassani Sadi, M., Sudarshan, C. & Wehn N.
Publié dans: Springer Journal of Design Automation for Embedded Systems (2023), 2024, ISSN 1572-8080
Éditeur: Springer
DOI: 10.1007/s10617-024-09282-2

Conference proceedings (8)

Energy Efficiency Impact of Processing in Memory: A Comprehensive Review of Workloads on the UPMEM Architecture

Auteurs: Yann Falevoz & Julien Legriel
Publié dans: Euro-Par 2023: Parallel Processing Workshops, 2024, ISSN 0302-9743
Éditeur: Springer Cham
DOI: 10.1007/978-3-031-48803-0_13

EC-NAS: Energy Consumption Aware Tabular Benchmarks for Neural Architecture Search

Auteurs: Pedram Bakhtiarifard; Christian Igel; Raghavendra Selvan
Publié dans: 2024, ISSN 2379-190X
Éditeur: IEEE
DOI: 10.1109/ICASSP48485.2024.10448303

Activation Compression of Graph Neural Networks Using Block-Wise Quantization with Improved Variance Minimization

Auteurs: Eliassen, Sebastian; Selvan, Raghavendra
Publié dans: Crossref, 2024, ISSN 2379-190X
Éditeur: IEEE
DOI: 10.48550/arxiv.2309.11856

Probabilistic feature matching for fast scalable visual prompting

Auteurs: Thomas Frick∗ , Cezary Skura∗ , Filip M. Janicki∗ , Roy Assaf , Niccolo Avogaro , Daniel Caraballo , Yagmur G. Cinar , Brown Ebouky , Ioana Giurgiu , Takayuki Katsuki , Piotr Kluska , Cristiano Malossi , Haoxiang Qiu , Tomoya Sakai , Florian Scheidegger
Publié dans: International Joint Conference for Artificial Intelligence (IJCAI 2024), August 2024, 2024, ISSN 0000-0000
Éditeur: IJCAI

Latent Inspector: An Interactive Tool for Probing Neural Network Behaviors Through Arbitrary Latent Activation.

Auteurs: Daniel Geißler, Bo Zhou, Paul Lukowicz.
Publié dans: International Joint Conference for Artificial Intelligence (IJCAI 2023), May 2023, 2023
Éditeur: Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence

CoSS: Co-optimizing Sensor and Sampling Rate for Data-Efficient AI in Human Activity Recognition

Auteurs: Mengxi Liu, Zimin Zhao, Daniel Geißler, Bo Zhou, Sungho Suh, Paul Lukowicz
Publié dans: AAAI 2024 Sustainable AI workshop, January 2024, 2024, ISSN 2331-8422
Éditeur: arXiv
DOI: 10.48550/arXiv.2401.05426

The Power of Training: How Different Neural Network Setups Influence the Energy Demand

Auteurs: Daniel Geißler, Bo Zhou, Mengxi Liu, Sungho Suh, Paul Lukowicz
Publié dans: AAAI 2024 Sustainable AI workshop, January 2024, 2024, ISSN 2331-8422
Éditeur: arXiv
DOI: 10.48550/arXiv.2401.01851Focustolearnmore

FieldHAR: A Fully Integrated End-to-End RTL Framework for Human Activity Recognition with Neural Networks from Heterogeneous Sensors

Auteurs: Mengxi Liu; Bo Zhou; Zimin Zhao; Hyeonseok Hong; Hyun Kim; Sungho Suh; Vitor Fortes Rey; Paul Lukowicz.
Publié dans: 2023, ISSN 2160-052X
Éditeur: IEEE
DOI: 10.1109/ASAP57973.2023.00029