ABPlanner

Code for our paper "An Adaptable Budget Planner for Enhancing Budget-Constrained Auto-Bidding in Online Display Advertising" in KDD 2025.

Requirements

python>=3.6
pytorch==2.3.1
argparse
tqdm
scipy

Usage

First, add the project path to the PYTHONPATH environment variable.

export PYTHONPATH=$PWD 

To train ABPlanner in the pure simulation environment with PID auto-bidder, run the following command:

python ABPlanner/main_pure_PID.py

To train ABPlanner in the pure simulation environment with USCB auto-bidder, first train the USCB agent by running the following command:

python PureSimEnv/USCBAgent/main.py --save_dir PureSimEnv/USCBAgent/result/

Then, run the following command to train ABPlanner:

python ABPlanner/main_pure_USCB.py

License

MIT