文档介绍：The first of these, Timesl, is the version used Introduction
to generate the table above. The second incorpo-
rates a modification similar to that applied to CF Consider the following prototype discrete opti-
at the price of introducing a loop, and the third mization monly known as the (un-
avoids the loop with the use of the Each operator. bounded) Knapsack Problem:
On the ten derivatives at a single point prob-
lem, the modified versions of Times were not an N
improvement, but in all other problems, the f(b):= max ~ w nx n (1)
Times2 function (with the loop) was consistently (xl .....XN) n = 1
faster than either of the other two, and Times3
appeared to be faster than Timesl when the data subject to:
was nested.
N
Average 10-1 10-25 25-1 25-25 E vn xn _< b (2)
(in seconds) n=l
Times I
x, •{0,I,2,...} (3)
Times 2
Times 3 for l _< n_< N
where {On}, {wn} and b are given positive integers.
putation of the first ten derivatives of For simplicity it is assumed that at least -
the test function at 100 points, which took sec- ponent of the vector u is equal to 2. If this is not
onds with the functions of [2], pleted in so, then set v÷1, v and w÷O, w.
just under five seconds using Times2 and the This problem can be interpreted as follows: We
modifie