MATLAB-Based Validation and Comparative Analysis of Quantum-Dot Transistor Compute-in-Memory Architecture for Neural Network Applications
Engineering, Technology & Applied Science Research · 06.06.2026 · 16 (3) · 35186-35191
DOI: 10.48084/etasr.17793 ↗Kurz gefasst
Fachartikel in der Zeitschrift Engineering, Technology & Applied Science Research, online erschienen am 06.06.2026. Verfasst von Abdulrahman Husawi. Die Arbeit nennt 21 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
This work provides independent validation and extended analysis of a Quantum-Dot Transistor (QDT) based Compute-in-Memory (CIM) architecture, without claiming silicon-level experimental verification. A comprehensive MATLAB-based simulation framework was developed to validate and analyze a previously proposed QDT-CIM architecture. The simulation implements device-level QDT models, circuit-level multiplier cells, and system-level memory arrays to perform matrix-vector multiplication operations essential for neural network inference. The validation study demonstrates strong agreement with the error rates reported in the original paper, with multiplication errors of approximately 0.1%, addition errors of 0.05%, and activation errors of 1.0%, yielding a total accumulated error of approximately 1.15%. The neural network classification experiments on synthetic MNIST-like data confirm the architectural advantages of QDT-based CIM, showing significant speedup over traditional von Neumann computing while maintaining competitive accuracy. Beyond replicating original results, this work contributes: (i) a reproducible modular simulation framework, (ii) systematic error-source decomposition identifying ADC quantization as the dominant error contributor, (iii) new analyses of device variability, IR-drop effects, and array scalability, and (iv) Quantization-Aware Training (QAT) achieving 96.8% accuracy on real MNIST data. Energy analysis estimates 0.5 pJ/MAC, while comparison with GPU implementations (~30 pJ/MAC) suggests potential improvement of up to 60×, although direct comparison requires consideration of technology node, precision, and workload differences. This work establishes a reproducible simulation framework for future research in neuromorphic computing systems.
Zitieren
APA
Husawi, A. (2026). MATLAB-Based Validation and Comparative Analysis of Quantum-Dot Transistor Compute-in-Memory Architecture for Neural Network Applications. Engineering, Technology & Applied Science Research, 16(3), 35186-35191. https://doi.org/10.48084/etasr.17793
Vancouver
Husawi A. MATLAB-Based Validation and Comparative Analysis of Quantum-Dot Transistor Compute-in-Memory Architecture for Neural Network Applications. Engineering, Technology & Applied Science Research. 2026;16(3):35186-35191. doi:10.48084/etasr.17793
BibTeX
@article{Husawi2026,
author = {Husawi, Abdulrahman},
title = {MATLAB-Based Validation and Comparative Analysis of Quantum-Dot Transistor Compute-in-Memory Architecture for Neural Network Applications},
journal = {Engineering, Technology & Applied Science Research},
year = {2026},
volume = {16},
number = {3},
pages = {35186--35191},
doi = {10.48084/etasr.17793}
}
EUVIMED