the official code for PNMTF-2D

• dataset
• model
• code
• experiment.ini
• dataset.ini

dataset dir: dataset_name/processed/，with pre-processed files text.pkl and label.pkl

• text.pkl: pre-processed documents
• label.pkl: the corresponding labels of each document（shape as [2,3,4,1,1,...,2,3,4]）

the tfidf.pkl and vocab.pkl will be generated by our code.

the parameter settings

the required information for each dataset

mpi4py，python -m pip install mip4py. Python version recommended : 3.8.5, some unexpected problem will happen with newer versions.

• running command：

  mpiexec -n n_threads python PNMTF-2D-V1.py --data_name classic4 --exp_ini super1-4_PNMTF-2D-V1_CLASSIC4 --pr n_rows --pc n_cols

  • main paras：
    • data_name exp_ini to specify the dataset and paras
    • pc、pr：the numbers of col and row threads, pr * pc = p
• command on Tianhe-2：

  yhrun -N 8 -n 192 -p bigdata python3 -u ./PNMTF-2D-V1.py --data_name classic4 --exp_ini super1-4_PNMTF-2D-V1_CLASSIC4 --pr 32 --pc 6

  • main paras：
    • data_name exp_ini to specify the dataset and paras
    • pc、pr：the numbers of col and row threads, pr * pc = p
    • -N：num of nodes；
    • -n：num of threads；
    • -p：computation region。