Temporal-Sequential Learning With a Brain-Inspired Spiking Neural Network and Its Application to Musical Memory

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• Existing models inadequately represent and memorize sequential information in the brain.
• Sequence learning is a critical cognitive function requiring robust memory and temporal encoding.

Purpose of the Study:

• To introduce a novel spiking neural network (SNN) model that addresses limitations in current sequence learning models.
• To incorporate psychological and neurobiological findings for a more comprehensive cognitive model.
• To validate the model's efficacy using musical memory as an application.

Main Methods:

• Developed a collaborative subnetwork architecture with biologically plausible neurons and dynamic synaptic plasticity (STDP).
• Integrated a cortical-striatal loop-inspired dependent timing module for temporal information encoding.
• Implemented goal-based and episodic retrieval mechanisms operating at multiple time scales.

Main Results:

• The SNN model demonstrated high accuracy in storing and recalling large volumes of melody data.
• The model successfully retrieved entire melodies from partial episodes and varying playback speeds.
• The model's architecture, including subnetworks and temporal encoding, proved effective for sequence learning.

Conclusions:

• The proposed SNN model offers a significant advancement in simulating brain sequence learning and memory.
• The model's ability to handle temporal information and diverse retrieval methods enhances its cognitive plausibility.
• This biologically inspired approach provides a powerful tool for understanding and modeling complex cognitive functions.