Stylistic Composition of Melodies Based on a Brain-Inspired Spiking Neural Network

Area of Science:

• Computational Neuroscience
• Artificial Intelligence
• Algorithmic Music Composition

Background:

• Current AI music composition lacks biological plausibility, differing significantly from human cognitive processes.
• Bridging the gap between artificial intelligence and human brain function is crucial for advancing generative models.

Purpose of the Study:

• To develop a biologically plausible spiking neural network for algorithmic music composition.
• To simulate brain structures and musical information processing for enhanced AI creativity.

Main Methods:

• A novel spiking neural network architecture inspired by brain regions (e.g., musical memory, knowledge learning).
• Utilized a hierarchical neural network for musical knowledge formulation.
• Employed biologically plausible neural models with spike-timing-dependent plasticity (STDP) and brain oscillation frequencies.

Main Results:

• The model successfully simulates cooperation between different brain regions for generating stylistic melodies.
• Demonstrated genre-based and composer-based melody composition through distinct neural circuits.
• Experimental results confirm the model's ability to compose melodies in various styles.

Conclusions:

• The proposed spiking neural network offers a biologically plausible approach to algorithmic composition.
• This model advances AI by integrating principles of neural computation and musical cognition.
• Future AI development can benefit from understanding and replicating brain-based generative processes.