[GAN-testing] Evaluation GAN's Predication Accuracy

[llmhyy]

Hi @lylytran

Still on the methods in Issue#101, you may kindly evaluate the accuracy of GAN for each decision node with both branches covered. For each such decision node, we can try to let GAN generate 100 data points for TRUE/FALSE label respectively. You may change the training sample size from 100, 200, ..., 1000.

As a result, you may kindly generate an EXCEL report indicating the accuracy to generate data points for TRUE/FALSE label under training sample size K (K=100, 200, ..., 1000).

Thanks a lot!

The following the target methods for testing in Math project.

===================================================== org.apache.commons.math.distribution.BinomialDistributionImpl.cumulativeProbability [L2T: 1, Ran: 0.77] org.apache.commons.math.distribution.PoissonDistributionImpl.probability [L2T: 1, Ran: 0.77] org.apache.commons.math.special.Gamma.regularizedGammaP (line 145) [L2T: 0.81, Ran: 0.77] org.apache.commons.math.special.Beta.regularizedBeta (line 118) [L2T: 0.81, Ran: 0.55] org.apache.commons.math.distribution.HypergeometricDistributionImpl.cumulativeProbability [L2T: 0.75, Randoop: 1.0] org.apache.commons.math.random.RandomDataImpl.nextPoisson [L2T: 0.75, Randoop: close to 1] org.apache.commons.math.fraction.Fraction.getReducedFraction [L2T: 1, Randoop: 0.5]

[llmhyy]

The generated data points need to follow the following template.

template.xlsx