Bioemus-IBEA-windows

IBEA version that could be run from a host PC, hopefully interfacing with more than one board. It does NOT contain the source C++ files to setup the FPGA, so it is not intended to be cloned on a board. It is just to configure the network.

Installation

To install the package create a conda environment and install dependences:

conda create -n IbeaBioemus
conda activate IbeaBioemus
conda install --file requirements.txt

Usage

To use the package you should follow these steps:

1. Open the mainIBEA.ipynb notebook.
2. Set the variable client_ip_address with the IP address of your FPGAs. (It can be a list of multiple addresses, parallel execution should work).
3. Set the path of the save_folder where saving the results, and the goal_path where taking the goals. The goals must be a .CSV file containing the files <names_of_metrics>.CSV, with the names in lower case.
4. Set the lbound and ubound as the lower and upper bounds of the parameters that IBEA uses for its search.
5. Set the hyperparameters of the IBEA algorithm: n_initial_individuals, simulation_time, n_offspring and max_generations are the most important ones. For a full description of the parameters look at the IBEA class in the ibea.py file.
6. Run the cell.

Results

The results are saved in the specified save_path. They consists in:

1. A data folder containing all the binary files.
2. A file Output.txt containing all the information about the run.
3. A file monitor.txt containig the best fitness value contained in each run.
4. a folder Best_configurations containing the configuration files of the individual with the highest area of the polar plot.

Plot

To plot the results use the file plot_figures.ipynb. You only need to insert the path where results are saved and the path where goals are saved.