Soyed

Dr.-Ing. Soyed Tuhin Ahmed

Short Bio

Soyed has received his bachelor's in Electrical and Electronics Engineering from American International University Bangladesh with Summa Cum Laude and University Gold Medal in 2016 (around a year earlier than the expected completion time). During his bachelor's, he was actively participating in IEEE region 10 and Engineering Student Associations of Bangladesh activities at his university. He has received his master's in Communication Engineering from Technische Universität München (TUM) in the fall 2019/2020 semester. During his master's degree, he was a research intern at the chair of design automation (Lehrstuhl für Entwurfsautomatisierung) under the supervision of Apl. Prof. Dr.-Ing. habil. Helmut Gräb (IEEE Fellow) and Robert Bosch Corporate Research in Renningen, Germany. Also, he has worked part-time at Infineon Technologies AG as a software engineer. In his master thesis, he has studied the soft error resiliency of peripheral modules of embedded SoC in safety-critical applications under the supervision of Prof. Ulf Schlichtmann and Daniel Müller-Gritschneder. He has joined the CDNC group of Professor Tahoori (IEEE Fellow) at KIT University, Karlsruhe, Germany in September 2020. His current research interests are neuromorphic computing, resilient hardware accelerator for machine learning, robust and accurate deep learning, hardware architectures for neural network applications, computation in memory, and emerging memory computing paradigms.

Publications

Journals
Soyed Tuhin Ahmed, Mehdi B. Tahoori
One-Shot Online Testing of Deep Neural Networks Based on Distribution Shift Detection
in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems (TCAD), DOI, PDF, Oct 2024.
Soyed Tuhin Ahmed, M Mayahinia, M Hefenbrock, C Münch, MB Tahoori
Design-Time Reference Current Generation for Robust Spintronic-Based Neuromorphic Architecture
in ACM Journal on Emerging Technologies in Computing Systems 20 (JETC), DOI, PDF, Jan 2024.
Ahmed, Soyed Tuhin; Danouchi, Kamal; Hefenbrock, Michael; Prenat, Guillaume; Anghel, Lorena; Tahoori, Mehdi B.
Algorithm-Hardware Co-Design for Uncertainty Estimation on Spintronic-Based Architectures
in ACM Transactions on Embedded Computing System (EsWeek Special Issue, Volume 22, Issue 5s), DOI, PDF, Oct 2023.
Soyed Tuhin Ahmed, Kamal Danouchi, Michael Hefenbrock, Guillaume Prenat, Lorena Anghel, Mehdi B. Tahoori
SpinBayes: Algorithm-Hardware Co-Design for Uncertainty Estimation Using Bayesian In-Memory Approximation on Spintronic-Based Architectures
in ACM Transactions on Embedded Computing Systems (Volume 22, Issue 5s) , DOI, PDF, Sep 2023.
Soyed Tuhin Ahmed and Mehdi B. Tahoori
Fault-tolerant Neuromorphic Computing with Memristors Using Functional ATPG for Efficient Re-calibration
in IEEE Design & Test (Volume 40, Issue 4), DOI, PDF, Aug 2023.
[Revised Publication for Top Picks from the 40th IEEE VTS 2022]
Soyed Tuhin Ahmed; Michael Hefenbrock; Christopher Münch; Mehdi B. Tahoori
NeuroScrub+: Mitigating Retention Faults Using Flexible Approximate Scrubbing in Neuromorphic Fabric Based on Resistive Memories
in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems (Volume 42, Issue 5), DOI, PDF, May 2023.
Soyed Tuhin Ahmed , Kamal Danouchi, Christopher Münch, Guillaume Prenat, Lorena Anghel, Mehdi B. Tahoori
SpinDrop: Dropout-Based Bayesian Binary Neural Networks with spintronic Implementation
in IEEE Journal on Emerging and Selected Topics in Circuits and Systems (Volume 13, Issue 1), DOI, PDF, Feb 2023.
Soyed Tuhin Ahmed, Michael Hefenbrock, Christopher Münch, Mehdi B. Tahoori
NeuroScrub+: Mitigating Retention Faults Using Flexible Approximate Scrubbing in Neuromorphic Fabric Based on Resistive Memories
in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2022.
Conferences
Soyed Tuhin Ahmed, Michael Hefenbrock, Mehdi B. Tahoori
Tiny Deep Ensemble: Uncertainty Estimation in Edge AI Accelerators via Ensembling Normalization Layers with Shared Weights
in ICCAD '24: Proceedings of the 43rd IEEE/ACM International Conference on Computer-Aided Design, DOI, PDF, Oct 2024.
Soyed Tuhin Ahmed; Surendra Hemaram; Mehdi B. Tahoori
NN-ECC: Embedding Error Correction Codes in Neural Network Weight Memories using Multi-task Learning
in IEEE 42nd VLSI Test Symposium (VTS), DOI, PDF, Apr 2024.
Soyed Tuhin Ahmed; Kamal Danouchi; Guillaume Prenat; Lorena Anghel; Mehdi B. Tahoori
NeuSpin: Design of a Reliable Edge Neuromorphic System Based on Spintronics for Green AI
in Design, Automation and Test in Europe (DATE), DOI, PDF, Mar 2024.
Soyed Tuhin Ahmed; Kamal Danouchi; Guillaume Prenat; Lorena Anghel; Mehdi B. Tahoori
Enhancing Reliability of Neural Networks at the Edge: Inverted Normalization with Stochastic Affine Transformations
in Design, Automation and Test in Europe (DATE), DOI, PDF, Mar 2024.
Soyed Tuhin Ahmed, Roman Rakhmatullin, Mehdi B. Tahoori
Online Fault-Tolerance for Memristive Neuromorphic Fabric Based on Local Approximation
in 28th IEEE European Test Symposium, DOI, PDF, May 2023.
Soyed Tuhin Ahmed, Kamal Danouchi, Michael Hefenbrock, Guillaume Prenat, Lorena Anghel, Mehdi B. Tahoori
Scalable Spintronics-based Bayesian Neural Network for Uncertainty Estimation
in Design, Automation and Test in Europe Conference, DOI, PDF, Apr 2023.
Soyed Tuhin Ahmed, Kamal Danouchi, Christopher Münch, Guillaume Prenat, Lorena Anghel, Mehdi B Tahoori
Binary Bayesian Neural Networks for Efficient Uncertainty Estimation Leveraging Inherent Stochasticity of Spintronic Devices
IEEE/ACM International Symposium on Nanoscale Architectures (NANOARCH), DOI, PDF, Dec 2022.
Soyed Tuhin Ahmed, Mehdi B. Tahoori
Compact Functional Test Generation for Memristive Deep Learning Implementations using Approximate Gradient Ranking
IEEE International Test Conference (ITC), DOI, PDF, Sep 2022.
Atousa Jafari; Mahta Mayahinia; Soyed Tuhin Ahmed; Christopher Münch; Mehdi B. Tahoori
MVSTT: A Multi-Value Computation-in-Memory based on Spin-Transfer Torque Memories
in 25th Euromicro Conference on Digital System Design (DSD), DOI, PDF, Aug 2022.
Soyed Tuhin Ahmed, Mahta Mayahinia, Michael Hefenbrock, Christopher Münch, and Mehdi B. Tahoori
Process and Runtime Variation Robustness for Spintronic-Based Neuromorphic Fabric
IEEE European Test Symposium (ETS), DOI, PDF, May 2022.
Soyed Tuhin Ahmed, Mehdi B. Tahoori
Fault-tolerant Neuromorphic Computing with Functional ATPG for Post-manufacturing Re-calibration
40th VLSI Test Symposium (VTS), DOI, PDF, Apr 2022.
Soyed Tuhin Ahmed; Michael Hefenbrock; Christopher Münch; Mehdi B. Tahoori
NeuroScrub: Mitigating Retention Failures Using Approximate Scrubbing in Neuromorphic Fabric Based on Resistive Memories
in IEEE European Test Symposium (ETS, DOI, PDF, May 2021.
Soyed Tuhin Ahmed*, Surendra Hemaram*, Mehdi B. Tahoori
Embedding Neural Network Parameters with Self Error Correcting Coding using Multi-task Learning
in 42nd VLSI Test Symposium (VTS), 2024.
arXiv
Soyed Tuhin Ahmed, Kamal Danouchi, Michael Hefenbrock, Guillaume Prenat, Lorena Anghel, Mehdi B. Tahoori
Scale-Dropout: Estimating Uncertainty in Deep Neural Networks Using Stochastic Scale
in arXiv, DOI, PDF, Nov 2023.
Soyed Tuhin Ahmed, K Danouchi, G Prenat, L Anghel, MB Tahoori
Concurrent Self-testing of Neural Networks Using Uncertainty Fingerprint
arXiv preprint arXiv:2401.01458, 2024.
Soyed Tuhin Ahmed, K Danouchi, G Prenat, L Anghel, MB Tahoori
Testing Spintronics Implemented Monte Carlo Dropout-Based Bayesian Neural Networks
arXiv preprint arXiv:2401.04744, 2024.
Ahmed, Soyed Tuhin, Mehdi B.
One-Shot Online Testing of Deep Neural Networks Based on Distribution Shift Detection
arXiv preprint arXiv:2305.09348, 2023.
Ahmed, Soyed Tuhin; Danouchi, Kamal; Hefenbrock, Michael; Prenat, Guillaume; Anghel, Lorena; Tahoori, Mehdi B.
Spatial-SpinDrop: Spatial Dropout-based Binary Bayesian Neural Network with Spintronics Implementation
arXiv preprint arXiv:2306.10185, 2023.
Other
Ahmed, Soyed Tuhin; Danouchi, Kamal; Münch, Christopher; Prenat, Guillaume; Anghel, Lorena; Tahoori, Mehdi B.
SpinDrop: Dropout-Based Bayesian Binary Neural Networks with spintronic Implementation
TechRxiv, DOI, 2022.
PhD-Forum
Soyed Tuhin Ahmed
Scalable and Efficient Methods for Uncertainty Estimation and Reduction in Deep Learning
in Design, Automation and Test in Europe (DATE), 2024.
Soyed Tuhin Ahmed
Reliable Memristive Neuromorphic In-Memory Computing: An Algorithm-Hardware Co-Design Approach
in Design, Automation and Test in Europe Conference, 2023.

Thesis and internship topics are available on

  1. Test and reliability aspect of Neuromorphic in-memory Computing.
  2. Model compression for deep learning applications
  3. Bayesian neural networks