README.md

CubeProbe

CubeProbe estimates the dtv distance of a sampler from uniform distribution and tests if the sampler is almost uniform or far from uniform. CubeProbe achieves the goal by implementing a SubCube Conditioning access to the sampler.

CubeProbe can work in two modes:

1. Estimator mode : In this mode CubeProbe returns the estimated total variation distance of the sampler from uniform distribution.
2. Tester mode : In this mode CubeProbe returns 'ACCEPT' if total variation distance is less than tolerance parameter and returns 'REJECT' if total variation distance is more than intolerance parameter.

Usage:

python CubeProbe.py --zeta ZETA --eta ETA --epsilon EPSILON --delta DELTA --sampler SAMPLERTYPE --seed SEED --thread THREAD --mode MODE input output

Here ZETA is the approximation parameter for the Estimator mode. ETA, EPSILON are the farness and closeness paramter in the Tester mode. DELTA is confidence parameter.

SAMPLERTYPE accepts 3 values:

• QuickSampler = 1
• STS = 2
• CMSGen = 3 SEED has been set to 420 in all the experiments. THREAD the is the number of threads used by CubeProbe. The arguement MODE takes two values:

1. est for Estimator mode.
2. test for Tester mode.

Corresponding paper available at here.