Learning-from-Label-Proportions

0. Please cite Clayton Scott and Jianxin Zhang, "Learning from Label Proportions: A Mutual Contamination Framework"
1. Dependencies:
   • for LMMCM:
     • download and install Anaconda
     • create an environment and install the dependencies: conda env create -f LLP.yml
     • activate the new environment: conda activate LLP
   • for InvCal and pSVM:
     • follow the readme file in the psvm folder
2. Generate LLP data:
   • open the folder LMMCM
   • usage: python make_data.py loader dataset_name path_to_save_mat path_to_save_binary lower_bound upper_bound train_size test_size [number_of_bags]
     • to re-generate the Adult and Magic data in the main paper and save them to the folder ./experiments/, run:
       • python make_data.py load_adult adult0 ./experiments/ ./experiments/ 0 0.5 8192 3000
       • python make_data.py load_adult adult1 ./experiments/ ./experiments/ 0.5 1 8192 3000
       • python make_data.py load_magic magic0 ./experiments/ ./experiments/ 0 0.5 5120 1400
       • python make_data.py load_magic magic1 ./experiments/ ./experiments/ 0.5 1 5120 1400
     • to re-generate the Adult and Magic data in the supplement and save them to the folder ./experiments/, run:
       • python make_data.py load_adult supp_adult0 ./experiments/ ./experiments/ 0 0.5 -1 3000 16
       • python make_data.py load_adult supp_adult1 ./experiments/ ./experiments/ 0.5 1 -1 3000 16
       • python make_data.py load_magic supp_magic0 ./experiments/ ./experiments/ 0 0.5 -1 1400 12
       • python make_data.py load_magic supp_magic1 ./experiments/ ./experiments/ 0.5 1 -1 1400 12
     • the created dataset will be named as [dataset name]_[bag size]_[trial id], trial id is an integer from 0 to 4
   • the program samples five subsets with repeatition for each bag size in [8, 16, 32, 64, 128, 256, 512] and saves the created dataset in a binary
   • the program saves a copy of the created dataset in .mat format
3. Reproduce the LMMCM resuls:
   • open the folder Learning-from-Label-Proportions
   • usage: python run_experiment.py data_path path_to_save_results
   • example: python run_experiment.py ./experiments/adult0_8_0 ./experiments/adult0_8_0_res
4. Reproduce InvCal and pSVM results:
   • open the folder pSVM
   • run InvCal: run_exp(@InvCal_cv, data_path, ouptput_path);
   • run pSVM run_exp(@alterPSVM_cv, data_path, ouptput_path);
   • the code is modified from Fexlix Yu's implementation of pSVM and InvCal