Related Experiment Video
Updated: Mar 16, 2026
07:34
A Simple Stimulatory Device for Evoking Point-like Tactile Stimuli: A Searchlight for LFP to Spike Transitions
Published on: March 25, 2014
10.3K
Supervised Learning in Spiking Neural Networks for Precise Temporal Encoding.
Brian Gardner1, André Grüning1
1Department of Computer Science, University of Surrey, Guildford, United Kingdom.
Plos One
|August 18, 2016
Summary
A new filtered error signal (FILT) learning rule enhances supervised learning in spiking neural networks for precise temporal coding. This biologically relevant rule improves pattern memorization and achieves efficient, online learning.
Area of Science:
- Computational Neuroscience
- Machine Learning
Background:
- Spiking neural networks (SNNs) offer advantages in processing information via precise spike timing.
- Developing biologically relevant supervised learning rules for SNNs remains a challenge.
Purpose of the Study:
- To investigate conditions for effective synaptic plasticity in SNNs for precise temporal coding.
- To compare two spike-based learning rules: instantaneous (INST) and filtered (FILT) error signals.
Main Methods:
- Analysis of synaptic plasticity conditions for temporal coding.
- Implementation and testing of INST and FILT learning rules in SNNs.
- Evaluation of temporal encoding precision and storage capacity (pattern memorization).
Main Results:
- The FILT rule demonstrates high performance in temporal encoding precision and pattern memorization.
- The FILT rule's error-filtering mechanism facilitates smooth convergence during learning.
- FILT rule shows efficiency in memorizing patterns, especially those with sub-millisecond precision.
Conclusions:
- The FILT rule offers a biologically relevant and theoretically grounded approach for supervised learning in SNNs.
- FILT rule performance is comparable to existing methods like the Chronotron rule.
- FILT rule's online implementation capability enhances its biological realism.
