Code, data, and figures for our paper.
All code for the paper appears in the code folder.
The data sources we use (listed below) tend to be quite large. We preprocessed a bunch of them and saved them in the data folder.
The taxi data set is particularly large. We load each individually and infer routes using Dijkstra's weighted shortest path algorithm. Then the total traffic on edge edge in the road network is saved to the flows folder.
Many of our models took substantial resources to train. We generally saved their weights in the saved_models folder.
We made one several pretty figures of Q values in Manhattan according to our deep Q GNN. You should check it out! Some of these figures are not up to date with the what we included in the paper.
• NYC Lion
If you have issues installing torch-sparse, torch-scatter and/or pyg-lib, check your torch version by running python, import torch, then torch.__version__. Once you exit the python environment, uninstall the current versions by running pip uninstall torch-scatter torch-sparse pyg-lib. Then, if your torch version is 2.0.1+cu117 run pip install torch-scatter torch-sparse pyg-lib -f https://data.pyg.org/whl/torch-2.0.1+cu117.html
https://stackoverflow.com/questions/60360750/lightgbm-classifier-with-gpu
Please reach out to us if you have any questions! :)