Real-Time TCAD: a new paradigm for TCAD in the artificial intelligence era

Myung, Sanghoon; Kim, Jin-Woo; Jeon, Yongwoo; Jang, Wonik; Huh, In; Kim, Jaemin; Han, Song-Yi; Baek, Kwan-Hyuck; Ryu, Jisu; Kim, Yoon-Suk; Doh, Ji-Seong; Kim, Jae-Ho; Jeong, Changwook; Kim, Dae Sin

doi:10.23919/sispad49475.2020.9241622

Cited by 13 publications

(15 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…10 and Table III, our evaluation framework leverages the MAPE as the primary performance metric, complemented by the coefficient of determination (R 2 ) metric. 39 This dualmetric approach facilitates a comprehensive assessment of the models' performance. The MAPE metric quantifies predictive error in of percentage error, providing a tangible measure of the model's precision and is given by:…”

Section: Resultsmentioning

confidence: 99%

Physics-Integrated Machine Learning for Efficient Design and Optimization of a Nanoscale Carbon Nanotube Field-Effect Transistor

Fan,

Low

2023

ECS J. Solid State Sci. Technol.

View full text Add to dashboard Cite

We propose a framework for optimizing the design of a carbon nanotube field-effect transistor (CNTFET) through the integration of device physics, machine learning (ML), and multi-objective optimization (MOO). First, we leverage the calibrated TCAD model based on experimental data to dissect the physical mechanisms of the CNTFET, gaining insights into its operational principles and unique physical properties. This model also serves as a foundation, enabling multi-scale performance evaluations essential for dataset construction. In the ML phase, a novel chain structure of support vector regression guided by a comprehensive statistical analysis of the design metrics yields a notable 97.35% accuracy without overfitting, even with limited data. The established ML model exhibits its competence in rapidly producing a global response surface for multi-scale CNTFET. Remarkably, an unusual trend within the CNTFET's behavior has been identified through this process, for which a physics explanation has been provided. We further compare shallow and deep learning-based TCAD digital twins for model selection guidance. Using the non-dominated sorted genetic algorithm-II (NSGA-II) in MOO, we harmonize metrics at both device and circuit levels, significantly reducing the design space. The closed-loop framework expedites the early-stage development of advanced transistors, overcoming the challenges posed by limited data.

show abstract

Section: Resultsmentioning

confidence: 99%

Physics-Integrated Machine Learning for Efficient Design and Optimization of a Nanoscale Carbon Nanotube Field-Effect Transistor

Fan,

Low

2023

ECS J. Solid State Sci. Technol.

View full text Add to dashboard Cite

show abstract

“…The previous studies [3]- [5] have only predicted the results. On the contrary, the proposed algorithm makes inputs and outputs of TCAD and RTT interchangeable.…”

Section: B Compatibility With Tcad Simulationmentioning

confidence: 88%

“…Last but not least, putting coordinates during performing convolution operations is very crucial for model performance. The previous studies [3], [12] have put relative coordinates in each CNN operation. On the other hand, in this study, the coordinates extracted by actual meshes must be inserted because the coordinates of uniformed-sized outputs could be different.…”

Section: B Rtt Process Modelmentioning

confidence: 99%

“…In the semiconductor field, two approaches have been mainly studied. The first one is to model the relationship between inputs and outputs using DL rather than using partial differential equations (PDE) [3]- [6]. The second one is to solve differential equation via DL, which provides the initial values for PDE [7] or regularizes itself to mimic the differential equation [8].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Restructuring TCAD System: Teaching Traditional TCAD New Tricks

Myung

Jang

Jin

et al. 2021

2021 IEEE International Electron Devices Meeting (IEDM)

Self Cite

View full text Add to dashboard Cite

Traditional TCAD simulation has succeeded in predicting and optimizing the device performance; however, it still faces a massive challenge -a high computational cost. There have been many attempts to replace TCAD with deep learning, but it has not yet been completely replaced. This paper presents a novel algorithm restructuring the traditional TCAD system. The proposed algorithmpredicts three-dimensional (3-D) TCAD simulation in real-time while capturing a variance, enables deep learning and TCAD to complement each other, and fully resolves convergence errors.

show abstract

“…The dataset consists of the scalar values as an input and images as an output. We used the RTT model (Myung et al, 2020(Myung et al, , 2021b to solve this problem.…”

Section: Real-world Problemmentioning

confidence: 99%

PAC-Net: A Model Pruning Approach to Inductive Transfer Learning

Myung¹,

Huh²,

Jang³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Inductive transfer learning aims to learn from a small amount of training data for the target task by utilizing a pre-trained model from the source task. Most strategies that involve large-scale deep learning models adopt initialization with the pretrained model and fine-tuning for the target task. However, when using over-parameterized models, we can often prune the model without sacrificing the accuracy of the source task. This motivates us to adopt model pruning for transfer learning with deep learning models. In this paper, we propose PAC-Net, a simple yet effective approach for transfer learning based on pruning. PAC-Net consists of three steps: Prune, Allocate, and Calibrate (PAC). The main idea behind these steps is to identify essential weights for the source task, fine-tune on the source task by updating the essential weights, and then calibrate on the target task by updating the remaining redundant weights. Under the various and extensive set of inductive transfer learning experiments, we show that our method achieves state-of-the-art performance by a large margin.

show abstract

Real-Time TCAD: a new paradigm for TCAD in the artificial intelligence era

Cited by 13 publications

References 11 publications

Physics-Integrated Machine Learning for Efficient Design and Optimization of a Nanoscale Carbon Nanotube Field-Effect Transistor

Physics-Integrated Machine Learning for Efficient Design and Optimization of a Nanoscale Carbon Nanotube Field-Effect Transistor

Restructuring TCAD System: Teaching Traditional TCAD New Tricks

PAC-Net: A Model Pruning Approach to Inductive Transfer Learning

Contact Info

Product

Resources

About