Boosting TalTech Capacity in Reliable and Efficient AI-Chip Design

Informations projet

TAICHIP

N° de convention de subvention: 101160182

DOI

10.3030/101160182

Date de signature de la CE 30 Avril 2024

Date de début 1 Septembre 2024

Date de fin 31 Août 2027

Financé au titre de

• Widening participation and spreading excellence

Coût total

Aucune donnée

Contribution de l’UE

€ 1 234 567,89

Coordonné par

TALLINNA TEHNIKAÜLIKOOL
Estonia

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 (5)

Plan for the dissemination and exploitation, including communication

Plan for the dissemination and exploitation, including communication (a plan, implementation and conclusive analysis). It reports on activities in tasks T5.1, T5.2 and T5.3.

Report on upskilling staff in research management

Report on TAICHIP research management capacity development related to the progress of tasks:T3.1 Upskilling staff in project proposal preparation and project managementT3.3 Upgrading research management/administration unit

Initial techniques for optimization and implementation of AI-chips

"Report on initial research results achieved in tasks T4.3 ""Efficiency optimisation of processor-accelerator pair"" and T4.4 ""Reliable hardware implementation of AI chips""."

Initial techniques for open AI-accelerator and processor architectures

"Report on initial research results achieved in tasks T4.1 ""Advanced AI-accelerator architectures"" and T4.2 ""Open processor architectures""."

Report on TAICHIP networking for excellence in research

Report on TAICHIP networking for excellence in research related to the progress of tasks:T1.1 Short-term scientific missions and virtual meetings within the consortiumT1.2 TAICHIP training eventsT1.3 Advanced technical tutorials by invited external expertsT1.4 Joint supervision of students and postdocs

Data Management Plan (1)

Research data management plan (DMP)

Research data management plan (DMP).

Publications

Peer reviewed articles (5)

An Efficient Architecture for Edge AI Federated Learning With Homomorphic Encryption

Auteurs: Dadmehr Rahbari, Masoud Daneshtalab, Maksim Jenihhin
Publié dans: IEEE Access, Numéro 13, 2025, ISSN 2169-3536
Éditeur: Institute of Electrical and Electronics Engineers (IEEE)
DOI: 10.1109/ACCESS.2025.3576689

AdAM: Adaptive Approximate Multiplier for Fault Tolerance in DNN Accelerators

Auteurs: Mahdi Taheri, Natalia Cherezova, Samira Nazari, Ali Azarpeyvand, Tara Ghasempouri, Masoud Daneshtalab, Jaan Raik, Maksim Jenihhin
Publié dans: IEEE Transactions on Device and Materials Reliability, Numéro 25, 2025, ISSN 1530-4388
Éditeur: Institute of Electrical and Electronics Engineers (IEEE)
DOI: 10.36227/TECHRXIV.171502587.72983622/V1

FORTALESA: Fault-tolerant reconfigurable systolic array for DNN inference

Auteurs: Natalia Cherezova, Artur Jutman, Maksim Jenihhin
Publié dans: Microprocessors and Microsystems, Numéro 119, 2025, ISSN 0141-9331
Éditeur: Elsevier BV
DOI: 10.1016/J.MICPRO.2025.105222

Special Session: In-Field ML-Assisted Intermittent Fault Localization and Management in RISC-V SoCs

Auteurs: Selg, Hardi; Shibin, Konstantin; Tsertov, Anton; Jenihhin, Maksim; Ellervee, Peeter; Raik, Jaan
Publié dans: 2024 IEEE International Symposium on Defect and Fault Tolerance in VLSI and Nanotechnology Systems (DFT), 2024, ISSN 2765-933X
Éditeur: IEEE
DOI: 10.1109/DFT63277.2024.10753541

AdAM: Adaptive Approximate Multiplier for Fault Tolerance in DNN Accelerators

Auteurs: Mahdi Taheri, Natalia Cherezova, Samira Nazari, Ali Azarpeyvand, Tara Ghasempouri, Masoud Daneshtalab, Jaan Raik, Maksim Jenihhin
Publié dans: IEEE Transactions on Device and Materials Reliability, Numéro 25, 2025, ISSN 1530-4388
Éditeur: Institute of Electrical and Electronics Engineers (IEEE)
DOI: 10.1109/TDMR.2024.3523386

Conference proceedings (9)

AxEnMULT: Design of an Efficient and Reliable Approximate Encoding-Based Multiplier

Auteurs: Ahsan Rafiq, Alberto Bosio, Salvatore Pappalardo, Maksim Jenihhin
Publié dans: 2025 IEEE Computer Society Annual Symposium on VLSI (ISVLSI), 2025
Éditeur: IEEE
DOI: 10.1109/ISVLSI65124.2025.11130248

Adaptive Fault Resilience for Early-Exit DNNs

Auteurs: Kodamanchili, Rama Mounika; Cherezova; Natalia; Taheri, Mahdi; Jenihhin, Maksim
Publié dans: IEEE Asian Test Symposium (ATS'25) (ATS-25)
DOI: 10.5281/ZENODO.18462542

Reliability-Aware Performance Optimization of DNN HW Accelerators Through Heterogeneous Quantization

Auteurs: Samira Nazari, Mahdi Taheri, Ali Azarpeyvand, Mohsen Afsharchi, Christian Herglotz, Maksim Jenihhin
Publié dans: 2025 IEEE 26th Latin American Test Symposium (LATS), 2025
Éditeur: IEEE
DOI: 10.1109/LATS65346.2025.10963948

FORTUNE: A Negative Memory Overhead Hardware-Agnostic Fault TOleRance TechniqUe in DNNs

Auteurs: Samira Nazari, Mahdi Taheri, Ali Azarpeyvand, Mohsen Afsharchi, Tara Ghasempouri, Christian Herglotz, Masoud Daneshtalab, Maksim Jenihhin
Publié dans: 2024
Éditeur: Institute of Electrical and Electronics Engineers (IEEE)
DOI: 10.36227/techrxiv.173143099.94285566/v1

Special Session: In-Field ML-Assisted Intermittent Fault Localization and Management in RISC-V SoCs

Auteurs: Hardi Selg, Konstantin Shibin, Anton Tsertov, Maksim Jenihhin, Peeter Ellervee, Jaan Raik
Publié dans: 2024 IEEE International Symposium on Defect and Fault Tolerance in VLSI and Nanotechnology Systems (DFT), 2024
Éditeur: IEEE
DOI: 10.1109/DFT63277.2024.10753541

FORTUNE: A Negative Memory Overhead Hardware-Agnostic Fault TOleRance TechniqUe in DNNs

Auteurs: Samira Nazari, Mahdi Taheri, Ali Azarpeyvand, Mohsen Afsharchi, Tara Ghasempouri, Christian Herglotz, Masoud Daneshtalab, Maksim Jenihhin
Publié dans: 2024 IEEE 33rd Asian Test Symposium (ATS), 2025
Éditeur: IEEE
DOI: 10.1109/ATS64447.2024.10915463

GENIE: GENetIc Algorithm-Based REliability Assessment Methodology for Deep Neural Networks

Auteurs: Samira Nazari, Mahdi Taheri, Ali Azarpeyvand, Mohsen Afsharchi, Christian Herglotz, Maksim Jenihhin
Publié dans: 2025 11th International Conference on Computing and Artificial Intelligence (ICCAI), 2025
Éditeur: IEEE
DOI: 10.1109/ICCAI66501.2025.00049

XMULT: An Energy-Efficient Design of Approximate Multiplier

Auteurs: Ahsan Rafiq, Maksim Jenihhin
Publié dans: 2025 55th Annual IEEE/IFIP International Conference on Dependable Systems and Networks Workshops (DSN-W), 2025
Éditeur: IEEE
DOI: 10.1109/DSN-W65791.2025.00047

RL-Agent-based Early-Exit DNN Architecture Search Framework

Auteurs: Mahdi Taheri, Parth Patne, Natalia Cherezova, Ali Mahani, Christian Herglotz, Maksim Jenihhin
Publié dans: 2025 IEEE 28th International Symposium on Design and Diagnostics of Electronic Circuits and Systems (DDECS), 2025
Éditeur: IEEE
DOI: 10.1109/DDECS63720.2025.11006795