Presentation | 2023-11-10 Examination of high high-precision device modeling methods Kengo Nakata, Takayuki Mori, Jiro Ida, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Neural network (NN) models have the advantage of high inference speed, but they are difficult to modeling. For this reason, a hybrid model of the Berkeley Short-channel IGFET Model (BSIM) and the NN model has recently been proposed. However, in this approach, the inference performance, which is the good point of NN, is rate-limited by BSIM. Therefore, we wondered if the NN model accuracy could be improved by creating a model for training, as in distillation. In this study, device modeling of MOSFETs was conducted using a linear regression model as a model for training NN, which can be trained at a lower cost and with higher accuracy than NN models, and it was found that the drain current variation (3σ) could be modeled with a high accuracy of 0.02%. By utilizing this model, we believe there is a possibility to develop a NN model with higher accuracy than the current one. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | neural network / SPICE / linear regression |
Paper # | SDM2023-71 |
Date of Issue | 2023-11-02 (SDM) |
Conference Information | |
Committee | SDM |
---|---|
Conference Date | 2023/11/9(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Process, Device, Circuit simulation, etc. |
Chair | Shunichiro Ohmi(Tokyo Inst. of Tech.) |
Vice Chair | Tatsuya Usami(Rapidus) |
Secretary | Tatsuya Usami(Tohoku Univ.) |
Assistant | Takuji Hosoi(Kwansei Gakuin Univ.) / Takuya Futase(Western Digital) |
Paper Information | |
Registration To | Technical Committee on Silicon Device and Materials |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Examination of high high-precision device modeling methods |
Sub Title (in English) | Comparison of Neural Networks and Linear Regression |
Keyword(1) | neural network |
Keyword(2) | SPICE |
Keyword(3) | linear regression |
1st Author's Name | Kengo Nakata |
1st Author's Affiliation | Kanazawa Institute of Technology(Kanazawa Inst. Tech.) |
2nd Author's Name | Takayuki Mori |
2nd Author's Affiliation | Kanazawa Institute of Technology(Kanazawa Inst. Tech.) |
3rd Author's Name | Jiro Ida |
3rd Author's Affiliation | Kanazawa Institute of Technology(Kanazawa Inst. Tech.) |
Date | 2023-11-10 |
Paper # | SDM2023-71 |
Volume (vol) | vol.123 |
Number (no) | SDM-250 |
Page | pp.pp.36-40(SDM), |
#Pages | 5 |
Date of Issue | 2023-11-02 (SDM) |