Master's Thesis - Matthew Judijanto - RWTH Aachen 2021
Title: Graph Neural Networks for Compact Representation for Job Shop Scheduling Problems: A Comparative Benchmark
- PyTorch Geometric
- PyTorch 1.7.0
Datasets of this thesis are generated based on:
- Normal datasets: https://www.scitepress.org/Link.aspx?doi=10.5220/0010202405890597
- L2D datasets: https://arxiv.org/abs/2010.12367
Note: Dataset for problem above the size of 10x10 are removed in PyG form due to GitHub 50mb file size limit.
-
Baseline: -- MLP
MLP.py
-- Ensemble BoostingRunEnsembleBoosting.py
-
GNN Models: -- GCN
GCN.py
-- GIN-0GIN.py
-- PNAPNA.py
-- EGCEGC.py
To run and train the models, there are 2 ways to do this. For a simple single run, you can use RunModel.py
. For running a K-Fold Cross Validation use cv_split_test.py
DataExtractor.py
DataExtractorNorm.py
DisjunctiveGraph.py
DisjunctiveGraph_v2.py
are used to convert the raw dataset from JSON to PyG graph data.
loadDataset.py
is to load the PyG graph datasets to be runned on the model