Doctoral Schools WUT

Search Engine for Promoters and Research Areas

Ph.D., D.Sc. Witold Pleskacz
Faculty of Electronics and Information Technology
Dane kontaktowe:
Gmach Elektroniki, p. 370
Repozytorium PW:

Wykaz obszarów badawczych:

Obszar badawczy Dziedzina naukowa

Building a semiconductor-device model for Monte Carlo simulation takes a large set of nominally identical structures to be electrically characterized. This can be impossible in the case of experimental devices or those working in non-standard conditions, e.g. at extreme temperatures. Generative models based on machine learning techniques may offer a solution by generating the required amount of synthetic data having statistical properties identical to the electrical responses of actual devices. Training a generative model, however, also takes a large training set of device measurements. A solution may lie in transfer learning, i.e. pretraining the model on large datasets of responses of electrically similar, easily available devices, followed by fine-tuning on a smaller set of target devices. Transfer learning in the context of generative models is a largely unexplored topic.

Automation, Electronics, Electrical Engineering and Space Technologies