18. We prove a stronger general conclusion: If real numbers a1,a2,⋯,an+1, and a1⩾a2⩾⋯⩾an,an+1=a1, then
k=1∑nf(ak+1)ak⩽k=1∑nf(ak)ak+1
When n=2, the proposition is obviously true (since it is an equality). Assume the proposition holds for n=m. Then for n=m+1, for b1⩾b2⩾⋯⩾bm+1,bm+2=b1, we have
∑k=1m+1f(bk+1)bk−∑k=1m+1f(bk)bk+1=∑k=1m−1f(bk+1)bk−∑k=1m−1f(bk)bk+1+f(bm+1)bm+f(b1)bm+1−f(bm)bm+1−f(bm+1)b1=[∑k=1m−1f(bk+1)bk+f(b1)bm−∑k=1m−1f(bk)bk+1−f(bm)b1]+[(b1−bm+1)f(bm)−(b1−bm)f(bm+1)−(bm−bm+1)−f(b1)]
By the assumption, the first polynomial is less than or equal to 0. By the convexity of f(x), the second polynomial is also less than or equal to 0. Therefore, when n=m+1, the proposition holds. In summary, this conclusion is true. When n=2003, it is the problem at hand.