Related Experiment Video
Updated: Jun 11, 2026
09:58
Syringe-injectable Mesh Electronics for Stable Chronic Rodent Electrophysiology
Published on: July 21, 2018
23.9K
Temporally Reconfigurable Reservoir Computing with Flexible Electrolyte-Gated TFTs for High-Performance Neuromorphic
Kang Hyun Lee1, Seohak Park2, Mingu Kang2
1Graduate School of Semiconductor Technology, Korea Advanced Institute of Science and Technology (KAIST), 291 Daehak-ro, Yuseong-gu, Daejeon, 34141, Republic of Korea.
Advanced Materials (Deerfield Beach, Fla.)
|September 17, 2025
Summary
This study presents a flexible neuromorphic system using novel thin-film transistors for advanced reservoir computing (RC). The system efficiently processes complex spatiotemporal data, achieving high accuracy in image classification tasks.
Area of Science:
- Neuromorphic Engineering
- Materials Science
- Computer Science
Background:
- Reservoir computing (RC) is a brain-inspired algorithm for spatiotemporal signal processing.
- Conventional RC systems struggle with diverse temporal scales and spatial complexities due to fixed dynamics.
Purpose of the Study:
- To develop a temporally reconfigurable RC system using high-performance ultrathin, flexible, all-solid-state electrolyte-gated thin-film transistors (UFLEX TFTs).
- To enable modulation of temporal dynamics for enhanced spatiotemporal signal processing.
- To demonstrate the system's capability in complex image classification tasks.
Main Methods:
- Fabrication of UFLEX TFTs with MoS2 channels and AlOx dielectrics.
- Integration of UFLEX TFTs into a reconfigurable RC system.
- Electrical characterization including on/off ratio, endurance, and variability.
- Evaluation of mechanical flexibility and performance after bending.
- Testing classification accuracy on CIFAR-10 object images and NIH chest X-ray images.
Main Results:
- UFLEX TFTs exhibit high performance: on/off ratio ≈10^7, endurance >2.5 × 10^4 pulses, and low variability.
- The flexible RC system maintains robust performance after bending.
- Achieved 90.3% accuracy for CIFAR-10 and 81.8% for NIH chest X-ray image classification.
- Demonstrated feature extraction across varied temporal and spatial scales.
Conclusions:
- The developed temporally reconfigurable RC system offers a flexible and efficient hardware solution.
- UFLEX TFTs enable dynamic modulation of temporal processing for neuromorphic applications.
- This work provides a foundation for advanced, flexible neuromorphic hardware systems.
Related Concept Videos
01:24Electrochemical Systems
Electrochemical systems provide a fascinating insight into the dynamic interplay of charged species within various phases. One notable example is the interaction between a membrane permeable to K⁺ ions but not to Cl⁻ ions, separating an aqueous KCl solution from pure water. As K⁺ ions diffuse through the membrane, they generate net charges on each phase, leading to a potential difference between them.Similarly, when a piece of Zn is immersed in an aqueous ZnSO₄ solution, the Zn metal, composed...
01:24Types of Reversible Electrodes
For electrode reversibility to be maintained, all the reactants and products involved in the half-reaction must be present at the electrode. There are several types of reversible electrodes (half-cells).In metal-metal-ion electrodes, a metal balances electrochemically with a solution of its own ions. Examples are Cu2+|Cu and Zn2+|Zn. Metals that react with the solvent, like group 1 and most group 2 metals, which react with water, and zinc, which reacts with aqueous acidic solutions, cannot be...