Advancing Compact Modeling of Electronic Devices: Machine Learning Approaches with Neural Networks, Mixture Density Networks, and Deep Symbolic Regression

This thesis pioneers the integration of deep learning techniques into the realm of compact modeling, presenting three distinct approaches that enhance the precision, efficiency, and adaptability of compact models for electronic devices. The first method introduces a Generalized Multilayer Perception Compact Model, leveraging the function approximation capabilities of neural networks through a multilayer perception (MLP) framework. This approach utilizes hyperband tuning to optimize network hyperparameters, demonstrating its effectiveness on a HfOx memristor and establishing a versatile modeling strategy for both single-state and multistate devices.The second approach explores the application of Mixture Density Networks (MDNs) to encapsulate the inherent stochasticity of electronic devices in compact models. By capturing complex probability distributions and employing an innovative sampling methodology based on inverse transform sampling, this method offers a more accurate representation of cycle-to-cycle variations in stochastic devices, as evidenced through experimental data from a heater cryotron.Finally, the thesis ventures into deep symbolic regression to construct compact models that facilitate faster convergence in circuit simulations. This method employs a transformer-based neural network architecture to derive equations directly from device datasets, significantly streamlining the model creation process and enhancing simulation efficiency. The applicability and benefits of this approach are exemplified through the development of a compact model for a graphene nanoribbon field-effect transistor (GNRFET).Collectively, these approaches embody a significant advancement in compact modeling, offering novel solutions that harness the power of deep learning to address the complexities and variabilities of modern electronic devices. This work not only contributes to the field of electronic device modeling but also opens new avenues for research in leveraging deep learning for enhanced computational simulations and model generalization.