The chronotron: a neuron that learns to fire temporally precise spike patterns

Area of Science:

• Computational Neuroscience
• Neural Networks
• Machine Learning

Background:

• Neurons process information via precisely timed spikes.
• Temporal coding is crucial for neural information processing.

Purpose of the Study:

• Introduce novel supervised learning rules for spiking neurons with temporal coding (chronotrons).
• Investigate the capacity of chronotrons to process and memorize temporally coded information.
• Enhance understanding of neural information processing and memory.

Main Methods:

• Developed two learning rules: E-learning (high memory capacity) and I-learning (biological plausibility).
• Utilized computer simulations with integrate-and-fire neurons.
• Analyzed synaptic changes proportional to synaptic currents for I-learning.

Main Results:

• Chronotrons accurately generate temporally precise output spikes with sub-millisecond precision.
• Demonstrated chronotrons' ability to classify inputs and reduce noise.
• Computed memory capacity bounds and explored parameter influences.

Conclusions:

• Chronotrons effectively model neurons encoding information in spike timing or phase.
• Firing one spike per cycle optimizes memory capacity in oscillatory networks.
• This work advances the understanding of temporal coding in neural systems.