A Spike-Driven Neuromorphic Framework for Asynchronous Population-Based Optimisation: Experiment Codes and Dataset

This repository contains datasets and scripts generated from experiments conducted using a neuromorphic optimisation framework. The materials are provided solely to support anonymous peer review, and the full experimental protocols and results are described in the corresponding manuscript. About This Repository This Zenodo repository serves exclusively for reproducibility, providing: Raw experimental data from all performed benchmark runs. Experimental configurations and parameter files. Scripts and processing routines used to generate the experimental figures and tables reported in the paper. The comprehensive experiments assess the functionality, scalability, and runtime performance of the NeurOptimiser framework across the BBOB test suite, utilising both linear and Izhikevich spiking neuron models. Nevertheless, the information provided here can be easily adapted to other neuromorphic optimisation algorithms and problems. What is inside this Zenodo Repository? This repository is organised into two main components: datasets and scripts. Datasets exconf.zip: YAML configuration files used to launch each batch of experiments. exdata-ioh.zip: Resulting plots from experiments conducted using the script exp_00-ioh.py, which implements the BBOB test suite from IOHexperimenter. exdata-coco.zip: Raw results dataset generated using the script exp_01-coco.py , along with the exp_01-coco-*.yaml configuration files from exconf.zip. ppdata-coco.zip: Postprocessed datasets generated by cocopp from COCO platform employing the raw results from exdata-coco.zip. exdata-time.zip: Raw data for timing analysis and runtime scalability evaluation, generated by exp_01-coco.py with the time_*.yaml configuration files from exconf.zip. exdata-ablation.zip: Database generated by submit_ablations.sh using Optuna and a post-processed ablation report (in HTML) generated with ablation_report.py. Scripts neuroptimiser.zip: A frozen version of the framework repository sufficient to reproduce the reported experiments is included. Example0.ipynb: Jupyter notebook showing how to implement the simplest optimisation procedure using the NeurOptimiser framework. Example1.ipynb: Jupyter notebook showing how to implement a neuroptimiser, using default parameters, to solve a BBOB problem from IOH. Example2.ipynb: Jupyter notebook showing how to implement a neuroptimiser, using different parameters, to solve a BBOB problem from COCO. exp_00-ioh.py: Script to run the experiments and generate the raw data and plots from experiments with IOHexperimenter. Results are saved in exdata-ioh.zip. exp_01-coco.py: Script to run the experiments and generate raw data and plots from experiments with COCO. Raw and post-processed results are saved in exdata-coco.zip and ppdata-coco.zip, respectively. This script requires YAML configuration files from exconf.zip to run the experiments. python exp_01-coco.py ./exconf/toy.yaml 1 1 # Args: <config_file> <num_batches> <batch> exp_02-time.ipynb: Jupyter notebook for timing analysis and runtime scalability evaluation. The raw data used in this notebook is saved in exdata-time.zip, which was also generated with the exp_01-coco.py and with the time_*.yamlconfiguration files in exconf.zip. exp_03-ablation.zip: This ZIP contains all the scripts created and utilised for conducting the ablation study. Related Publication The experiments associated with this dataset are described in scientific publications that cite it. Please refer to the corresponding publication for detailed methodology and results.