Large Deviation Analysis for Hypothesis Testing for Score based Models

[arXiv] This is an implementation of Large Deviation Analysis for Hypothesis Testing for Score based Models

Requirements

See requirements.txt

Instructions

• Experimental control are configured in config.yml
• Use make.sh to generate run script with make.py
• Use make.py to generate exp script to scripts
• Use make_dataset.py to prepare datasets
• Use process.py to process exp results
• Experimental setup are listed in make.py
• Hyperparameters can be found in config.yml and process_control() of module/hyper.py

Examples

• Test of Multivariate Normal (MVN) distribution with pertubation $\sigma_{ptb} = 0.02$ on $\mu$ for theoretical limit

  python test_ht.py --control_name MVN_mvn_lrt-t_0.02-0.0_1
  

• Test of KDDCUP dataset (KDDCUP99) with "back" adversarial network traffic on $W$ of Gauss-Benoulli RBM for empirical limit ($N=10$)

  python test_ht.py --control_name KDDCUP99_rbm_hst-t_back_1_10
  

Results

• Large deviation analysis of likelihood-based and sore-based hypothesis testing for multivariate normal distribution with perturbation on $\mu$ and $\sigma_{ptb} = 0.02$.

Acknowledgements

Enmao Diao
Taposh Banerjee
Vahid Tarokh