For specific Python packages, a requirements file has been provided at 'requirements.txt'. Assuming Pythonis installed, run the following command to install all necessary dependencies:
pip install -r requirements.txt
All result files were available upon request due to size limitations.
To use the project, you can use the run_all.py file in the terminal for both top-level direction for complete functionality, or you can test each experiment individually for specific tasks. All experiments and their steps are organized as shown in the structure tree below:
python run_all.py --experiment all --step all
python run_all.py --experiment exp1 --step all
python run_all.py --experiment exp2 --step all
python run_all.py --experiment exp3 --step all
python run_all.py --experiment exp1 --step Step1
python run_all.py --experiment exp1 --step Step2
python run_all.py --experiment exp1 --step Step3
To run other experiments, simply replace exp1 with the desired experiment name, such as exp2 or exp3. To view Step 2 or Step 3 of one experiment, you should first run the previous step(s); otherwise, you won't be able to directly access the results of the prior step locally.
The dataset used in this project is substantial, and certain steps may require extended processing time.
For instance, on a 2 GHz Quad-Core Intel Core i5 with 16 GB 3733 MHz LPDDR4X:
Step 2 of Experiment 2 takes more than 12 hours to complete.
Step 2 of Experiment 3 takes approximately 36 hours to complete.
It is strongly recommended to run this code on a High-Performance Computing (HPC) cluster.