RISE Lab has been awarded a 3 year NSF grant to study how the interaction between long-term and short-term memory enables different energy efficient computing architecture. Specifically, we are interested in pursuing a variant of Ferroelectric Field Effect Transistor, we named DFeFET, where the drain terminal is engineered to allow mixed ionic-electronic conduction. This would enable short term plasticity in the drain current. The interaction between the short-term plasticity can interact with the long-term plasticity from the ferroelectric gate through the modulated threshold voltage. We aim to demonstrate that this new device technology can potentially revolutionize spiking neural networks and physical reservoir computing. Our ultimate target is to study the impact of utilizing short-term memory dynamics to more mainstream deep neural networks, as multiple theoretical papers predicted its positive impact. Even the pioneers of AI, Hopfield and Hinton also spoke about our lack of understanding in the effect of forgetting in brain (hence, short-term memory) with respect to accelerating deep neural network at a low energy expense. If you are curious to know more, please reach out to the PI.