A frugal Spiking Neural Network for unsupervised multivariate temporal pattern classification and multichannel spike...

Area of Science:

• Computational Neuroscience
• Machine Learning
• Signal Processing

Background:

• Advanced neural interfaces generate large volumes of continuous data requiring efficient processing algorithms.
• Current methods for analyzing neural data can be computationally intensive and may not be suitable for real-time applications.

Purpose of the Study:

• To introduce a frugal, generic, single-layer Spiking Neural Network (SNN) for unsupervised identification and classification of multivariate temporal patterns.
• To enable automatic, real-time pattern recognition in high-dimensional neural data streams.

Main Methods:

• Development of a single-layer Spiking Neural Network (SNN) architecture.
• Validation using simulated multivariate data, Mel Cepstral representations of speech, and multichannel neural recordings.
• Testing for online-compatible, unsupervised classification of action potentials in spike sorting datasets.

Main Results:

• The proposed SNN effectively identified and classified multivariate temporal patterns in simulated and real neural data.
• The SNN demonstrated successful unsupervised classification of action potentials in spike sorting tasks.
• The approach proved effective in an online-compatible mode.

Conclusions:

• This frugal SNN architecture offers a viable solution for automatic, unsupervised, real-time pattern recognition in neural data.
• The findings suggest potential for embedding these SNNs into ultra-low power hardware for applications like active neural implants.