Evolutionary Multi-task Optimization with Adaptive Knowledge Transfer

This repository implements AEMTO and three other EMTO comparison methods with adaptive knowledge transfer (in main), i.e. MaTDE, SBO, MFEA2.

Envs

The following software environments are required. The code is only tested on linux OS. If you want to run on other OS, some equivalent environments should be set properly.

cmake >= 3.2
g++ >= 6.4.0; use c++11 standard
python >= 3.0

Build and Run with default arguments

git clone https://github.com/haoxuhao/AEMTO.git
cd AEMTO
pip install -r requirements.txt
mkdir build && cd build && cmake .. && make -j
cd ../bin
./AEMTO # && ./MATDE && ./SBO && ./MFEA2

The detailed results are recorded in bin/Results/*/*.json. Json results file can be opened directly by common text editor or parsed by most programming languages.

Others

File bin/run.sh lists some useful information to do experiments. Directory bin/scripts contains the python scripts to compare and visualize the results of different algorithms.

Acknowledgements

Open source softwares in oss:

and the base version of MATDE from author of paper MaTEA.