cerc-io/solidity
16
0
Fork 0
mirror of https://github.com/ethereum/solidity synced 2023-10-03 13:03:40 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

[yul-phaser] SimulationRNG: Rewrite uniformInt() and binomialInt() tests to rely on variance and expected value

Browse Source
This commit is contained in:
Kamil Śliwak 2020-02-14 14:53:37 +01:00
parent db140a667a
commit a8fa332a9c
1 changed files with 28 additions and 46 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

74
test/yulPhaser/SimulationRNG.cpp
View File

@ -15,6 +15,8 @@
along with solidity. If not, see <http://www.gnu.org/licenses/>.
*/
#include <test/yulPhaser/Common.h>
#include <tools/yulPhaser/SimulationRNG.h>
#include <boost/test/unit_test.hpp>
@ -31,36 +33,22 @@ BOOST_AUTO_TEST_SUITE(RandomTest)
BOOST_AUTO_TEST_CASE(uniformInt_returns_different_values_when_called_multiple_times)
{
constexpr uint32_t numSamples = 1000;
constexpr uint32_t numOutcomes = 100;
SimulationRNG::reset(1);
constexpr size_t numSamples = 1000;
constexpr uint32_t minValue = 50;
constexpr uint32_t maxValue = 80;
constexpr double relativeTolerance = 0.05;
vector<uint32_t> samples1;
vector<uint32_t> samples2;
// For uniform distribution from range a..b: EX = (a + b) / 2, VarX = ((b - a + 1)^2 - 1) / 12
constexpr double expectedValue = (minValue + maxValue) / 2.0;
constexpr double variance = ((maxValue - minValue + 1) * (maxValue - minValue + 1) - 1) / 12.0;
vector<uint32_t> samples;
for (uint32_t i = 0; i < numSamples; ++i)
{
samples1.push_back(SimulationRNG::uniformInt(0, numOutcomes - 1));
samples2.push_back(SimulationRNG::uniformInt(0, numOutcomes - 1));
}
samples.push_back(SimulationRNG::uniformInt(minValue, maxValue));
vector<uint32_t> counts1(numOutcomes, 0);
vector<uint32_t> counts2(numOutcomes, 0);
for (uint32_t i = 0; i < numSamples; ++i)
{
++counts1[samples1[i]];
++counts2[samples2[i]];
}
// This test rules out not only the possibility that the two sequences are the same but also
// that they're just different permutations of the same values. The test is probabilistic so
// it's technically possible for it to fail even if generator is good but the probability is
// so low that it would happen on average once very 10^125 billion years if you repeated it
// every second. The chance is much lower than 1 in 1000^100 / 100!.
//
// This does not really guarantee that the generated numbers have the right distribution or
// or that they don't come in long, repeating sequences but the implementation is very simple
// (it just calls a generator from boost) so our goal here is just to make sure it's used
// properly and we're not getting something totally non-random, e.g. the same number every time.
BOOST_TEST(counts1 != counts2);
BOOST_TEST(abs(mean(samples) - expectedValue) < expectedValue * relativeTolerance);
BOOST_TEST(abs(meanSquaredError(samples, expectedValue) - variance) < variance * relativeTolerance);
}
BOOST_AUTO_TEST_CASE(uniformInt_can_be_reset)
@ -96,30 +84,24 @@ BOOST_AUTO_TEST_CASE(uniformInt_can_be_reset)
BOOST_TEST(samples3 != samples4);
}
BOOST_AUTO_TEST_CASE(binomialInt_returns_different_values_when_called_multiple_times)
BOOST_AUTO_TEST_CASE(binomialInt_should_produce_samples_with_right_expected_value_and_variance)
{
constexpr uint32_t numSamples = 1000;
SimulationRNG::reset(1);
constexpr size_t numSamples = 1000;
constexpr uint32_t numTrials = 100;
constexpr double successProbability = 0.6;
constexpr double successProbability = 0.2;
constexpr double relativeTolerance = 0.05;
vector<uint32_t> samples1;
vector<uint32_t> samples2;
// For binomial distribution with n trials and success probability p: EX = np, VarX = np(1 - p)
constexpr double expectedValue = numTrials * successProbability;
constexpr double variance = numTrials * successProbability * (1 - successProbability);
vector<uint32_t> samples;
for (uint32_t i = 0; i < numSamples; ++i)
{
samples1.push_back(SimulationRNG::binomialInt(numTrials, successProbability));
samples2.push_back(SimulationRNG::binomialInt(numTrials, successProbability));
}
samples.push_back(SimulationRNG::binomialInt(numTrials, successProbability));
vector<uint32_t> counts1(numTrials, 0);
vector<uint32_t> counts2(numTrials, 0);
for (uint32_t i = 0; i < numSamples; ++i)
{
++counts1[samples1[i]];
++counts2[samples2[i]];
}
// See remark for uniformInt() above. Same applies here.
BOOST_TEST(counts1 != counts2);
BOOST_TEST(abs(mean(samples) - expectedValue) < expectedValue * relativeTolerance);
BOOST_TEST(abs(meanSquaredError(samples, expectedValue) - variance) < variance * relativeTolerance);
}
BOOST_AUTO_TEST_CASE(binomialInt_can_be_reset)