/** * \file RandomExact.cpp * \brief Using %RandomLib to generate exact random results * * Compile/link with, e.g.,\n * g++ -I../include -O2 -funroll-loops * -o RandomExact RandomExact.cpp ../src/Random.cpp\n * ./RandomExact * * See \ref otherdist, for more information. * * Copyright (c) Charles Karney (2006-2012) and licensed * under the MIT/X11 License. For more information, see * http://randomlib.sourceforge.net/ **********************************************************************/ #include #include #include #include #include #include #include #include #include #include #include int main(int, char**) { // Create r with a random seed RandomLib::Random r; r.Reseed(); std::cout << "Using " << r.Name() << "\n" << "with seed " << r.SeedString() << "\n\n"; { std::cout << "Sampling exactly from the normal distribution. First number is\n" << "in binary with ... indicating an infinite sequence of random\n" << "bits. Second number gives the corresponding interval. Third\n" << "number is the result of filling in the missing bits and rounding\n" << "exactly to the nearest representable double.\n"; const int bits = 1; RandomLib::ExactNormal ndist; long long num = 20000000ll; long long bitcount = 0; int numprint = 16; for (long long i = 0; i < num; ++i) { long long k = r.Count(); RandomLib::RandomNumber x = ndist(r); // Sample bitcount += r.Count() - k; if (i < numprint) { std::pair z = x.Range(); std::cout << x << " = " // Print in binary with ellipsis << "(" << z.first << "," << z.second << ")"; // Print range double v = x.Value(r); // Round exactly to nearest double std::cout << " = " << v << "\n"; } else if (i == numprint) std::cout << std::flush; } std::cout << "Number of bits needed to obtain the binary representation averaged\n" << "over " << num << " samples = " << bitcount/double(num) << "\n\n"; } { std::cout << "Sampling exactly from exp(-x). First number is in binary with\n" << "... indicating an infinite sequence of random bits. Second\n" << "number gives the corresponding interval. Third number is the\n" << "result of filling in the missing bits and rounding exactly to\n" << "the nearest representable double.\n"; const int bits = 1; RandomLib::ExactExponential edist; long long num = 50000000ll; long long bitcount = 0; int numprint = 16; for (long long i = 0; i < num; ++i) { long long k = r.Count(); RandomLib::RandomNumber x = edist(r); // Sample bitcount += r.Count() - k; if (i < numprint) { std::pair z = x.Range(); std::cout << x << " = " // Print in binary with ellipsis << "(" << z.first << "," << z.second << ")"; // Print range double v = x.Value(r); // Round exactly to nearest double std::cout << " = " << v << "\n"; } else if (i == numprint) std::cout << std::flush; } std::cout << "Number of bits needed to obtain the binary representation averaged\n" << "over " << num << " samples = " << bitcount/double(num) << "\n\n"; } { std::cout << "Sampling exactly from the discrete normal distribution with\n" << "sigma = 7 and mu = 1/2.\n"; RandomLib::DiscreteNormal gdist(7,1,1,2); long long num = 50000000ll; long long count = r.Count(); int numprint = 16; for (long long i = 0; i < num; ++i) { int k = gdist(r); // Sample if (i < numprint) std::cout << k << " "; else if (i == numprint) std::cout << std::endl; } count = r.Count() - count; std::cout << "Number of random variates needed averaged\n" << "over " << num << " samples = " << count/double(num) << "\n\n"; } { std::cout << "Sampling exactly from the discrete normal distribution with\n" << "sigma = 1024 and mu = 1/7. First result printed is a uniform\n" << "range (with covers a power of two). The second number is the\n" << "result of sampling additional bits within that range to obtain\n" << "a definite result.\n"; RandomLib::DiscreteNormalAlt gdist(1024,1,1,7); long long num = 20000000ll; long long count = r.Count(); long long entropy = 0; int numprint = 16; for (long long i = 0; i < num; ++i) { RandomLib::UniformInteger u = gdist(r); entropy += u.Entropy(); if (i < numprint) std::cout << u << " = "; int k = u(r); if (i < numprint) std::cout << k << "\n"; else if (i == numprint) std::cout << std::flush; } count = r.Count() - count; std::cout << "Number of random bits needed for full result (for range) averaged\n" << "over " << num << " samples = " << count/double(num) << " (" << (count - entropy)/double(num) << ")\n\n"; } { std::cout << "Random bits with 1 occurring with probability 1/pi exactly:\n"; long long num = 100000000ll; int numprint = 72; RandomLib::InversePiProb pp; long long nbits = 0; long long k = r.Count(); for (long long i = 0; i < num; ++i) { bool b = pp(r); nbits += int(b); if (i < numprint) std::cout << int(b); else if (i == numprint) std::cout << "..." << std::flush; } std::cout << "\n"; std::cout << "Frequency of 1 averaged over " << num << " samples = 1/" << double(num)/nbits << "\n" << "bits/sample = " << (r.Count() - k)/double(num) << "\n\n"; } { std::cout << "Random bits with 1 occurring with probability 1/e exactly:\n"; long long num = 200000000ll; int numprint = 72; RandomLib::InverseEProb ep; long long nbits = 0; long long k = r.Count(); for (long long i = 0; i < num; ++i) { bool b = ep(r); nbits += int(b); if (i < numprint) std::cout << int(b); else if (i == numprint) std::cout << "..." << std::flush; } std::cout << "\n"; std::cout << "Frequency of 1 averaged over " << num << " samples = 1/" << double(num)/nbits << "\n" << "bits/sample = " << (r.Count() - k)/double(num) << "\n"; } return 0; }