CMAS Lab - Prior Knowledge Accelerated Transfer Learning (PKI-TL) for

Journal article

Asha Kumari Jakhar, Surila Guglani, A. Dasgupta, Sourajeet Roy
2023 IEEE 32nd Conference on Electrical Performance of Electronic Packaging and Systems (EPEPS), 2023

Semantic Scholar DOI

Cite

APA Click to copy
Jakhar, A. K., Guglani, S., Dasgupta, A., & Roy, S. (2023). Prior Knowledge Accelerated Transfer Learning (PKI-TL) for Machine Learning Assisted Uncertainty Quantification of MLGNR Interconnect Networks. 2023 IEEE 32nd Conference on Electrical Performance of Electronic Packaging and Systems (EPEPS).

Chicago/Turabian Click to copy
Jakhar, Asha Kumari, Surila Guglani, A. Dasgupta, and Sourajeet Roy. “Prior Knowledge Accelerated Transfer Learning (PKI-TL) for Machine Learning Assisted Uncertainty Quantification of MLGNR Interconnect Networks.” 2023 IEEE 32nd Conference on Electrical Performance of Electronic Packaging and Systems (EPEPS) (2023).

MLA Click to copy
Jakhar, Asha Kumari, et al. “Prior Knowledge Accelerated Transfer Learning (PKI-TL) for Machine Learning Assisted Uncertainty Quantification of MLGNR Interconnect Networks.” 2023 IEEE 32nd Conference on Electrical Performance of Electronic Packaging and Systems (EPEPS), 2023.

BibTeX Click to copy

@article{asha2023a,
  title = {Prior Knowledge Accelerated Transfer Learning (PKI-TL) for Machine Learning Assisted Uncertainty Quantification of MLGNR Interconnect Networks},
  year = {2023},
  journal = {2023 IEEE 32nd Conference on Electrical Performance of Electronic Packaging and Systems (EPEPS)},
  author = {Jakhar, Asha Kumari and Guglani, Surila and Dasgupta, A. and Roy, Sourajeet}
}

Abstract

In this paper, an algorithm to combine the distinct advantages of knowledge-based training and transfer learning has been developed for the fast artificial neural network (ANN) assisted uncertainty quantification of on-chip multi-layered graphene nanoribbon (MLGNR) interconnect networks. In particular, the proposed algorithm enables different modes of information such as the values of the weights and bias terms and the predicted responses from a pre-trained secondary ANN to guide the highly data-efficient training of the primary ANN. The goal of the primary ANN is to develop a parametric model of the MLGNR interconnect responses that can be used in a Monte Carlo framework for fast uncertainty quantification.