Resource Selection in Cognitive Networks With Spiking Neural Networks

Area of Science:

• Computer Science
• Artificial Intelligence
• Neuroscience

Background:

• Cognitive networking aims to dynamically optimize network resource allocation.
• Low-power neuromorphic chips require efficient control mechanisms.
• Spiking neural networks (SNNs) offer a biologically inspired, energy-efficient computation model.

Purpose of the Study:

• To explore the feasibility of an SNN-based cognitive network controller (CNC).
• To design a CNC capable of dynamic resource optimization for recurrent network tasks.
• To evaluate the CNC's performance in a simulated space communication environment.

Main Methods:

• Developed a CNC utilizing SNN principles for resource selection.
• Implemented a spike-based coding strategy for action decisions.
• Introduced a learning algorithm and synapse regulation method.
• Simulated the CNC in a multichannel space communication link scenario.

Main Results:

• The proposed CNC successfully optimized average file transfer time.
• Achieved objective across a broad range of offered loads in a dynamic environment.
• Demonstrated effectiveness compared to conventional methods.
• Analyzed the impact of learning and space protocol parameters.

Conclusions:

• The SNN-based CNC is a feasible approach for cognitive networking.
• The CNC shows promise for optimizing resource selection in challenging, time-limited network environments.
• This work facilitates the development of novel cognitive networking applications for neuromorphic hardware.