PyNCS: a microkernel for high-level definition and configuration of neuromorphic electronic systems

Area of Science:

• Neuroscience
• Computer Engineering
• Software Development

Background:

• Neuromorphic hardware enables advanced sensory-motor systems and spiking neural networks.
• Current software interfaces are often monolithic and hardware-specific, hindering modularity and rapid development.
• A need exists for flexible, portable, and expandable software frameworks for neuromorphic systems.

Purpose of the Study:

• To introduce PyNCS, an open-source Python front-end for defining neural network models.
• To facilitate interfacing neuromorphic hardware with workstations through Python APIs.
• To promote modularity, portability, and expandability in neuromorphic software development.

Main Methods:

• Developed PyNCS as a modular framework with Python Application Programming Interfaces (APIs).
• Designed a high-level front-end for defining neural network models.
• Integrated tools for creating, monitoring, and analyzing spiking data.
• Demonstrated functionality with a neuromorphic vision sensor and a cognitive decision-making task.

Main Results:

• PyNCS successfully interfaces with diverse neuromorphic hardware setups.
• The framework supports the definition and analysis of spiking neural network models.
• Case studies validated PyNCS with an event-based vision sensor and a cognitive task.
• PyNCS accelerates the development of hybrid software/hardware neuromorphic systems.

Conclusions:

• PyNCS provides a flexible and modular solution for interacting with neuromorphic hardware.
• The framework's design promotes rapid evolution and reconfigurability of neuromorphic systems.
• PyNCS is poised to accelerate research and development in the field of neuromorphic computing.