Abstract
Analytical models have been generally used to describe the resistive switching (RS) behavior of resistive switching memory (RSM). Recently, the conduction filament (CF) growth models have been used to capture the device characteristics of a wide range of RSM devices. However, the assumptions used to model CF growth reduce the accuracy of real-world device characteristics. Machine learning holds the promise of modeling CF growth by learning the RSM device parameters with excellent accuracy via examples, bypassing the need to use analytical models. Here, we sidestep the accuracy issues by directly learning the device-compliance-current (CC) and aluminum-nitridebased-RSM device parameter values relationship in a simple way and with high accuracy for various device samples. We perform the first calculation with a machinelearned device parameter on aluminum-nitride-based RSM devices and are able to predict the CC value for filament growth using only a few RSM device parameters. The learning model now allows the construction of accurate RSM device parameters for realistic device modeling.