This repository contains the Python 3 implementation of SAEL. SAEL is based on the Situation-Aware Fear Leaning model (see the paper 'A Situation-Aware Fear Learning (SAFEL) model for robots' by Rizzi Raymundo, Caroline & Johnson, Colin & Fabris, Fabio & Vargas, Patricia. (2016) Neurocomputing. 221 pages 32-47. DOI: [https://doi.org/10.1016/j.neucom.2016.09.035])

It also contains the testing code, datasets used, and the summary results generated by the resultsProcesso.py script.

The full results are not included as they take up too much space. To recreate them you would need to run the tester.py script. Be warned this takes a long time (~1 week on an i7 CPU with 3 parallel threads) and several gigs of available memory. If you want to atempt it I would recoment you should adjust the maxConcurrency parameter in the script to match as many threads as you want your computer to dedicate to running the individual tests. Prediction results will not be exactly the same for the FrequencyPredictor but both tree predictors should give exact same prediction resulting in identical preformance scores as are in the summary files.

The python files with names starting in lower case are scripts, the ones with upper case names ending in Utils pull together several helper functions, and the remaining upper case named files are classes.

The main classes matching original SAFEL functionality are the SituationSegmenter class based on the Hippocampus module and the TreePredictor class based on the Working memory module.

There are two datasets in the Repository that were used for the testing. They are in the folders starting with SitSeq. These were generated with the dataset-generator available in https://github.com/fjpavm/dataset-generator