Overview

Brains-Py is an advanced Python package designed as a comprehensive software framework for implementing deep learning-based surrogate models of dopant network processing units (DNPUs) and for conducting in-depth analyses of their behaviour. By utilising surrogate models, Brains-Py facilitates the investigation of DNPU dynamics without the necessity of physical experimentation, thereby significantly conserving both time and resources.

Our result

Brains-Py is an advanced Python package designed as a comprehensive software framework for implementing deep learning-based surrogate models of dopant network processing units (DNPUs) and for conducting in-depth analyses of their behaviour. By utilising surrogate models, Brains-Py facilitates the investigation of DNPU dynamics without the necessity of physical experimentation, thereby significantly conserving both time and resources. Additionally, this package supports the evaluation of complex, scaled-up architectures composed of multiple DNPUs and enables the simulation of these architectures in conjunction with other components represented by artificial neural networks. Exclusively developed by the University of Twente, Brains-Py has been made publicly available on GitHub and is documented in the Journal of Open Source Software. 

Needs and target audiences

The target market for the solution includes academic researchers and industrial developers who work with dopant network processing units (DNPU) or similar technologies. These users benefit from a generic framework that can be adapted to various materials and different design aspects.

Current stage

The ideal scenario for the market is the establishment of a streamlined, resource-efficient methodology for investigating and optimising DNPU architectures. Researchers and developers require the ability to simulate complex systems without the need for substantial investments in physical prototypes. A

Unique value proposition

Brains-Py is a cutting-edge solution for simulating dopant network processing units (DNPUs), meeting the needs of academic and industrial researchers for efficient DNPU analysis. By eliminating physical experimentation, it saves significant time and costs and enables the evaluation of complex architectures, making it essential for advancing technologies like edge AI.

Keywords

• brain-inspired computing
• physical-substrate computing
• in-memory computing
• unconventional computing
• neuromorphic computing

Research project acronyms

• HYBRAIN

Research projects

• Hybrid electronic-photonic architectures for brain-inspired computing

Project identifiers

• 101046878

Contributors

• UNIVERSITEIT TWENTE

Source: Horizon Results Platform