Related Experiment Video
Updated: Aug 6, 2026
08:07
Assembly and Characterization of Biomolecular Memristors Consisting of Ion Channel-doped Lipid Membranes
Published on: March 9, 2019
Noise-Tunable Memristor Enabling Programmable Probabilistic Neurons for Frequency-Selective Time-Series Signal
Do Hoon Kim1, Seoeun Jang2, Hakseung Rhee1
1Department of Materials Science and Engineering, KAIST, Daejeon, Republic of Korea.
Advanced Materials (Deerfield Beach, Fla.)
|August 5, 2026
Summary
Memristors can now act as tunable entropy sources for probabilistic computing. This research introduces a novel neuron using memristor noise for adaptive, energy-efficient time-series encoding across diverse frequencies.
Area of Science:
- Materials Science
- Neuroscience
- Computer Engineering
Background:
- Memristors offer tunable resistance for memory and computing.
- Memristor instability enables entropy generation for security and probabilistic applications.
- Coupling tunable resistance and entropy in memristors is an underexplored area.
Purpose of the Study:
- To propose a spiking-rate-programmable probabilistic neuron.
- To utilize the tunable noise characteristics of a Ru/TaOx/Pt memristor.
- To explore frequency-selective, frequency-domain probabilistic neural encoding.
Main Methods:
- Developed a probabilistic neuron using a Ru/TaOx/Pt memristor.
- Exploited memristor's varying conduction mechanisms and noise behaviors across resistance states.
- Integrated probabilistic neurons into a network for time-series data classification.
Main Results:
- Demonstrated frequency-selective, frequency-domain probabilistic neural encoding.
- Achieved ~95% classification performance on both low-frequency (UCI HAR) and high-frequency (Audio MNIST) data.
- Showcased network adaptability to signals across a wide frequency range (0.4 Hz to 8 kHz).
Conclusions:
- Memristor noise can be harnessed for advanced neural encoding.
- Proposed neurons enable compact, adaptive, and energy-efficient time-series processing.
- Highlights a new paradigm for leveraging memristor properties in neuromorphic systems.
