Java内部实现：标准正态分布。即：Polar Method方法
public double nextDouble() {
return (((long)(next(26)) << 27) + next(27))
/ (double)(1L << 53);
}
private double nextNextGaussian;
private boolean haveNextNextGaussian = false;
synchronized public double nextGaussian() {
// See Knuth, ACP, Section 3.4.1 Algorithm C.
if (haveNextNextGaussian) {
haveNextNextGaussian = false;
return nextNextGaussian;
} else {
double v1, v2, s;
do {
v1 = 2 * nextDouble() - 1; // between -1 and 1
v2 = 2 * nextDouble() - 1; // between -1 and 1
s = v1 * v1 + v2 * v2;
} while (s >= 1 || s == 0);
double multiplier = StrictMath.sqrt(-2 * StrictMath.log(s)/s);
nextNextGaussian = v2 * multiplier;
haveNextNextGaussian = true;
return v1 * multiplier;
}
}