Classification of radio signals on a neuromorphic chip in space

This repository contains the source code for my master thesis project conducted at the Institute of Neuroinformatics in the fall semester of 2018. A reservoir computing approach was taken to implement a recurrent neural network for automatic radio signal modulation recognition on neuromorphic hardware for space applications. The simulations are developed using the Brian2 simulator and the Teili neuromorphic toolkit to mimic actual hardware as close as possible.

Organization

The repository is structured as follows:

• notebooks: Juypter notebooks documenting various steps of my research
• thesis: PDF version of the thesis and the presentation
• utils: library of various methods for data preprocessing, plotting and reservoir computing developed during the project
• spiking_radio_reservoir.py: main code base for the simulations
• bayesian_optimization_process.py: hyperparameter tuning for the reservoir