Vine_MCTS

Code for the paper

Chang, B., Pan, S. & Joe, H.. (2019). Vine copula structure learning via Monte Carlo tree search. International Conference on Artificial Intelligence and Statistics.

 

Install Python packages

Using pip:

pip install -r requirements.txt

See python-igraph website for detailed information on installing python-igraph package.

Usage

Generate correlation matrices

Use the following code to generate the correlation matrices, which are stored as .csv files in data/corr folder and used as input. For example, the following code will generate data/corr/abalone.csv.

cd data/raw_data/
python abalone.py
cd ../..

Customized input csv files can be created and saved in data/corr folder. The first line of the the csv files is the number of samples; the following lines are the correlation matrix. See data/corr/abalone.csv as an example.

Run MCTS

python main.py -f abalone -ntrunc 4,5 -max_iter 5000

Other arguments:

usage: main.py [-h] [-f FILE_PREFIX] [-nw NUM_WORKER] [-fpu FPU] [-pb PB]
               [-log_freq LOG_FREQ] [-ntrunc NTRUNC] [-seed SEED]
               [-max_iter MAX_ITER]

optional arguments:
  -h, --help          show this help message and exit
  -f FILE_PREFIX      File prefix
  -nw NUM_WORKER      Number of workers to run in parallel (default: 1)
  -fpu FPU            First play urgency (default: 1.0)
  -pb PB              Progressive bias (default: 0.1)
  -log_freq LOG_FREQ  Log Frequency (default: 100)
  -ntrunc NTRUNC      A list of truncation level. For example, "2,3,4". By
                      default, levels from 1 to d-1.
  -seed SEED          Seed (default: 1)
  -max_iter MAX_ITER  Maximum number of iterations (default: 5000)

FILE_PREFIX is the filename of the input csv file. Results are saved in output folder.

Interactive Jupyter Notebook example

Install Jupyter Notebook and run the following:

jupyter notebook misc/visualization.ipynb

BibTeX

@inproceedings{chang2019vine, 
  title={Vine copula structure learning via {M}onte {C}arlo tree search}, 
  author={Chang, Bo and Pan, Shenyi and Joe, Harry}, 
  booktitle={International Conference on Artificial Intelligence and Statistics}, 
  year={2019} 
}