EUVIMEDEUVIMED entdecken
Crossref

Spiking Super-Tsetlin: Superconducting Tsetlin Machines With Spiking Neurons

Ran Cheng, Christoph Kirst, Dilip Vasudevan

IEEE Transactions on Applied Superconductivity · 2026-08 · 36 (5) · 1301508-1301508

DOI: 10.1109/tasc.2026.3675576 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Applied Superconductivity, 2026-08. Verfasst von Ran Cheng und 2 weiteren Autor:innen. Die Arbeit nennt 38 Literaturangaben. Gefördert von: DOE Office of Advanced Scientific Computing Research.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The Tsetlin machine (TM) is a state-of-the-art machine learning approach where an ensemble of subclassifiers called clauses represents propositional logic formulas that are applied to the input sample in binary representation to describe or classify data. Super-Tsetlin is an implementation of the TM using superconducting rapid single-flux quantum technology. The current design of Super-Tsetlin relies on superconducting logic circuits, where the complexity of the clock network and pipeline depth limit the power and speed performance, especially for large-scale TMs, and its simplified learning scheme suffers from overtraining issues. In this work, we introduce a spiking neuron-based Super TM design to improve the performance and speed, as well as reduce the power consumption compared to the digital implementation of Super-Tsetlin. We estimated the major computational components to be at least four times more power-efficient and one to five times faster than the original Super-Tsetlin. In addition, we introduce a threshold circuit to prevent overtraining by comparing the current clause summation with a user-set target value $T$ and subsequently adjusting the probabilities of feedback signals to TAs during the training process. Moreover, the random number generator circuits, which play a key role in the randomness of TMs, are optimized to be more flexible in implementing the hyperparameter. We demonstrate the performance of the proposed systems through learning the modified national institute of standards and technology (MNIST) handwritten digit recognition task.

Förderung

  • DOE Office of Advanced Scientific Computing Research

Zitieren

APA

Cheng, R., Kirst, C., & Vasudevan, D. (2026). Spiking Super-Tsetlin: Superconducting Tsetlin Machines With Spiking Neurons. IEEE Transactions on Applied Superconductivity, 36(5), 1301508-1301508. https://doi.org/10.1109/tasc.2026.3675576

Vancouver

Cheng R, Kirst C, Vasudevan D. Spiking Super-Tsetlin: Superconducting Tsetlin Machines With Spiking Neurons. IEEE Transactions on Applied Superconductivity. 2026;36(5):1301508-1301508. doi:10.1109/tasc.2026.3675576

BibTeX
@article{Cheng2026,
  author = {Cheng, Ran and Kirst, Christoph and Vasudevan, Dilip},
  title = {Spiking Super-Tsetlin: Superconducting Tsetlin Machines With Spiking Neurons},
  journal = {IEEE Transactions on Applied Superconductivity},
  year = {2026},
  volume = {36},
  number = {5},
  pages = {1301508--1301508},
  doi = {10.1109/tasc.2026.3675576}
}
Kostenlos starten →