Low-Voltage Organic Transistor-Based Reservoir Computing for Multitask Applications.

Area of Science:

• Neuromorphic Engineering
• Organic Electronics
• Computational Neuroscience

Background:

• Physical reservoirs, especially transistor-based ones, show promise for temporal tasks due to nonlinear dynamics.
• Existing systems often require task-specific configurations, limiting reusability.

Purpose of the Study:

• To develop and validate a unified reservoir computing system using a low-voltage organic field-effect transistor (OFET).
• To demonstrate the system's capability for diverse temporal signal recognition tasks using a consistent protocol.

Main Methods:

• Implemented a reservoir computing system with a single low-voltage OFET.
• Developed a device-specific spike current map for reusable hardware pipelining.
• Preprocessed diverse inputs into a unified spike protocol for projection into high-dimensional states.

Main Results:

• Validated the OFET-based system on static image and dynamic gesture classification tasks.
• Achieved high accuracy across various gesture input modalities.
• Demonstrated consistent performance using the reusable spike current map.

Conclusions:

• The OFET-based reservoir computing system shows significant potential for neuromorphic gesture classification.
• A unified protocol and single-device architecture can effectively process diverse temporal signals.
• This approach offers a promising direction for transistor-based physical reservoirs.