1、在matlab中通过eps函数来求解某数的浮点相对精度，下面给出示例代码e1=epse2=eps('double')e3=eps('single')e4=eps(1/2)e5=eps(1)e6=eps(2)e7=eps(realmax)e8=eps(0)e9=eps(single(1/2))e10=eps(single(1))e11=eps(single(2))e12=eps(realmax('single'))e13=eps(single(0))
2、eps函数的一个用法是在需要考虑数字计算精度对结果的影响时，比如matlab中自带的求矩阵的秩的r锾攒揉敫ank函数默认返回的是奇异值大于max(size(A))*eps(norm(A))的个数，比如下面代码执行的结果为U=[ -0.379274725817546 0.794481008161344 0.040314689302254 0.472573101068419 -0.548121117865938 -0.158656907114637 0.785308487819421 -0.240170366475522 -0.701132103179362 -0.096930884878865 -0.615158038749589 -0.347273328675907 -0.253234469211253 -0.578128287711604 -0.056972731609046 0.773559367084354];V=[-0.395899761267122 0.174787331274778 -0.893374215953781 -0.120810918891644 -0.579336710586230 -0.801660599750723 0.083470911224132 0.121416084411623 -0.442564352640226 0.452790460112304 0.183006779832243 0.752081186800241 -0.558362918331707 0.348956259242386 0.401775603768836 -0.636424971841809];A=U*diag([10 20 30 1e-15])*V';def_tol=max(size(A))*eps(norm(A))rA1=rank(A)rA2=rank(A,1e-16)
