Litcius/Paper detail

Reservoir Computing System With HZO/Si FeFETs in Parallel Configuration: Experimental Demonstration of Speech Classification

Eishin Nako, Kasidit Toprasertpong, Ryosho Nakane, Mitsuru Takenaka, Shinichi Takagi

2023IEEE Transactions on Electron Devices19 citationsDOI

Abstract

We study a reservoir computing (RC) system with ferroelectric field-effect transistors (FeFETs) in a parallel configuration and develop various schemes in a speech classification task. Experimental drain-source, and substrate output currents of a FeFET are used for temporal reservoir state vectors in response to a time-series input signal at a corresponding frequency channel and their different characteristics accelerate the information extraction capability to effectively enhance the performance. Adjustable weights in the readout part are trained by Ridge regression. Finally, we achieved the highest classification accuracy of 98.1%. Our systematic approaches find important knowledge toward the system design establishment of FeFET-based RC for versatile application.

Topics & Concepts

Reservoir computingRidgeComputer scienceTransistorSIGNAL (programming language)Channel (broadcasting)Electronic engineeringSubstrate (aquarium)Field (mathematics)Task (project management)EngineeringArtificial intelligenceElectrical engineeringArtificial neural networkTelecommunicationsGeologyMathematicsVoltagePaleontologySystems engineeringPure mathematicsRecurrent neural networkProgramming languageOceanographyNeural Networks and Reservoir ComputingAdvanced Memory and Neural ComputingFerroelectric and Negative Capacitance Devices
Reservoir Computing System With HZO/Si FeFETs in Parallel Configuration: Experimental Demonstration of Speech Classification | Litcius