Achieving reinforcement learning in a three-active-terminal neuromorphic device based on a 2D vdW ferroelectric material

Feng Guo; Weng Fu Io; Zhaoying Dang; Ran Ding; Sin-Yi Pang; Yuqian Zhao; Jianhua Hao

doi:10.1039/d3mh00714f

Achieving reinforcement learning in a three-active-terminal neuromorphic device based on a 2D vdW ferroelectric material

Mater Horiz. 2023 Aug 29;10(9):3719-3728. doi: 10.1039/d3mh00714f.

Authors

Feng Guo^{1

2}, Weng Fu Io¹, Zhaoying Dang¹, Ran Ding¹, Sin-Yi Pang¹, Yuqian Zhao¹, Jianhua Hao^{1

2

3}

Affiliations

¹ Department of Applied Physics, The Hong Kong Polytechnic University, Hong Kong 999077, P. R. China. jh.hao@polyu.edu.hk.
² The Hong Kong Polytechnic University Shenzhen Research Institute, Shenzhen 518057, P. R. China.
³ Photonics Research Institute and Research Institute for Smart Energy, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong 999077, P. R. China.

PMID: 37403831
DOI: 10.1039/d3mh00714f

Abstract

Currently, for most three-terminal neuromorphic devices, only the gate terminal is active. The inadequate modes and freedom of modulation in such devices greatly hinder the implementation of complex neural behaviors and brain-like thinking strategies in hardware systems. Taking advantage of the unique feature of co-existing in-plane (IP) and out-of-plane (OOP) ferroelectricity in two-dimensional (2D) ferroelectric α-In₂Se₃, we construct a three-active-terminal neuromorphic device where any terminal can modulate the conductance state. Based on the co-operation mode, controlling food intake as a complex nervous system-level behavior is achieved to carry out positive and negative feedback. Specifically, reinforcement learning as a brain-like thinking strategy is implemented due to the coupling between polarizations in different directions. Compared to the single modulation mode, the chance of the agent successfully obtaining the reward in the Markov decision process is increased from 68% to 82% under the co-operation mode through the coupling effect between IP and OOP ferroelectricity in 2D α-In₂Se₃ layers. Our work demonstrates the practicability of three-active-terminal neuromorphic devices in handling complex tasks and advances a significant step towards implementing brain-like learning strategies based on neuromorphic devices for dealing with real-world challenges.