Self-Adaptation Spiking Neural Membrane Systems with Neuromodulators.

Area of Science:

• Computational Neuroscience
• Biologically Inspired Computing

Background:

• Spiking neural P systems (SN PS) are third-generation spiking neural networks (SNNs) for distributed computation.
• SN PS lack mechanisms to model neuromodulators, which influence synaptic plasticity in biological systems.

Purpose of the Study:

• Introduce a novel self-adaptive spiking neural P system with neuromodulators (SSNN PS).
• Enhance computational control in SN PS by incorporating neuromodulator-regulated self-adapting weights.

Main Methods:

• Neuromodulators are modeled as resources consumed by rules in a new postsynaptic membrane computational unit.
• The postsynaptic membrane features self-adapting weights regulated by neuromodulators, reflecting interneuronal connection intensity.
• Demonstrate Turing universality of SSNN PS for number generation and acceptance.

Main Results:

• SSNN PS exhibit enhanced control over the computational process.
• Turing universality was proven for SSNN PS.
• An SSNN PS model achieved 91.71% accuracy on UTKFace and 87.83% on FairFace for gender recognition, outperforming comparative methods.

Conclusions:

• SSNN PS offer a more biologically plausible model for SNNs by including neuromodulators.
• The self-adaptive mechanism improves computational accuracy, particularly in pattern recognition tasks.
• SSNN PS show significant potential for advanced artificial intelligence applications.