Skip to content

pandat8/ML4LB

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

<<<<<<< HEAD

Code and data for learning to search in local branching

Installation

  1. Please install Python 3.7, PyscipOPt 3.1.1, and SCIP 7.01 in your own computer environment

  2. install package ecole according to the following instructions:

    • go into the folder 'ecole'
    • install the package according to 'installation.rst'

Running the experiments

Only Computing and Plot the evaluation results in the paper (e.g. primal integral, primal gap)

compute_evaluation_results.py

Evaluating the trained model on 5 datasets by your own machine

# evaluate Algorithm lb-baseline, lb-sr, lb-srm, 
evaluation_regression_k_prime.py

# evaluate Algorithm lb-rl, lb-srmrl
evaluation_reinforce4lb.py

# compute and plot the evaluation results
compute_evaluation_results.py

Train your own regression model, RL model, and then evaluating them by your own machine

# train regression models
train_regression.py

# train RL models
train_reinforce4lb.py

# evaluate Algorithm lb-sr, lb-srm by your own regression model, evaluate lb-baseline 
evaluation_regression_k_prime.py --regression_model_path='path to your own model trained by mixed dataset' # after training, you can select the models from '.results/saved_models/regression/' folder 

# evaluate Algorithm lb-rl, lb-srmrl
evaluation_reinforce4lb.py --rl_model_path='path to your own model' # after training, you can select your models from '.results/saved_models/rl/reinforce/setcovering/' folder

# compute and plot the evaluation results
compute_evaluation_results.py

======= boost the priximity search of local branching algorithm with ml techniques.

Update README.md

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published