SlowElectronics

Concept

SlowElectronics indicates electronic technology suited for efficiently processing slowly varying signals (i.e. signals with relatively large timescales). Such signals are typically found in biological phenomena and human activities. Advanced electronics seeking overly speedy signal processing often encounters a difficulty in efficiently handling slowly varying signals. We are going to develop SlowElectronics towards realization of extremely efficient neuromorphic hardware devices by integrating methodologies in material science, electrical engineering, information science, and neuroscience. This concept was proposed in the JST CREST Project (Platform for real-time learning at the edge with spiking neural networks, 2019-2024, Grant No. JPMJCR19K2) directed by Dr. Isao H. Inoue at AIST.

Slow signals

Examples of target signals of SlowElectronics are shown below.

Reservoir computing for SlowElectronics

Reservoir computing is one of the machine learning frameworks promising for SlowElectronics. We have demonstrated that reservoir computing systems are suited for dealing with slowly varying time series data in individual studies. This site collects the information on the program codes developed in these studies.

Data	Task	Repository/URL	Model(Algorithm)	Reference
Handwriting Triangles (two persons)	Anomaly detection (for Authentication)	Folder	(1) SNN-reservoir (FORCE1)	Inoue et al., IEEE Symposium on VLSI Technology and Circuits, 2023
Handwriting Triangles (two persons)	Anomaly detection (for Authentication)	Folder	(1) ESN2 (LR3) (2) LSTM4 (BPTT5)	---
Human Blood Pressure	Classification (for Authentication)	Repository	(1) ESN2 (LR3) (2) Bidirectional-ESN (LR3)	Li et al., ICANN, 2023
UCR Anomaly Archive (A collection of 250 univariate time series collected in human medicine, biology, meteorology and industry)	Anomaly detection	Repository	(1) ESN2 (LR3) (2) MDRS6 (RLS7)	Tamura et al., TechRxiv, 2023
Human Blood Glucose (A Study to Assess Continuous Glucose Sensor Profiles in Healthy Non-Diabetic Participants Aged <7 Years)	Prediction	Repository	(1)ESN2 (RLS7)	Pati et al., Commun. Mater. 2024

Footnotes

1. FORCE learning (Sussillo and Abbott, 2009) ↩

2. Echo State Network (Jaeger, 2001) paper pdf ↩ ↩² ↩³ ↩⁴

3. Linear/Ridge Regression ↩ ↩² ↩³ ↩⁴

4. Long Short-Term Memory (Hochreiter and Schmidhuber, 1997) ↩

5. Backpropagation Through Time (Werbos et al., 1990) ↩

6. Mahalanobis Distance of Reservoir States (Tamura et al., 2023) ↩

7. Recursive Least Square (Jaeger, 2003) ↩ ↩²