Physical Reservoir Based on a Leaky-FeFET Using the Temporal Memory Effect
Gyusoup Lee, Changyeon Kang, Seongho Kim, Youngkeun Park, Eui Joong Shin, Byung Jin Cho
Abstract
In this work, a leaky-Ferroelectric Field Effect Transistor (FeFET) neuron is introduced as a physical reservoir in a reservoir computing scheme. Compared to a conventional FeFET reservoir control sample, which did not show leaky behavior, the proposed leaky-FeFET neuron-based physical reservoir exhibited 78.6% and 62.9% improvements in memory capacity for Short Term Memory (STM) and Parity Check (PC) tasks, respectively. The improvements are attributed to the temporal memory effect induced by the leaky-integrating neuronal behavior, which originates from the retention degradation of the FeFET.
Topics & Concepts
Reservoir computingTransistorComputer scienceField-effect transistorDynamic random-access memoryMaterials scienceOptoelectronicsElectronic engineeringElectrical engineeringVoltageEngineeringSemiconductor memoryComputer hardwareArtificial neural networkArtificial intelligenceRecurrent neural networkNeural Networks and Reservoir ComputingAdvanced Memory and Neural ComputingFerroelectric and Negative Capacitance Devices