《算法导论》

5.1-1

Show that the assumption that we are always able to determine which candidate is best in line 4 of procedure \text{HIRE-ASSISTANT}HIRE-ASSISTANT implies that we know a total order on the ranks of the candidates.

A total order is a partial order that is a total relation (\forall a, b \in A:aRb \text{ or } bRa)(∀a,b∈A:aRb or bRa). A relation is a partial order if it is reflexive, antisymmetric and transitive.

Assume that the relation is good or better.

• Reflexive: This is a bit trivial, but everybody is as good or better as themselves.
• Transitive: If AA is better than BB and BB is better than CC, then AA is better than CC.
• Antisymmetric: If AA is better than BB, then BB is not better than AA.

So far we have a partial order.

Since we assume we can compare any two candidates, then comparison must be a total relation and thus we have a total order.

5.1-2 \star⋆

Describe an implementation of the procedure \text{RANDOM}(a, b)RANDOM(a,b) that only makes calls to \text{RANDOM}(0, 1)RANDOM(0,1). What is the expected running time of your procedure, as a function of aa and bb?

As (b - a)(b−a) could be any number, we need at least \lceil \lg(b - a) \rceil⌈lg(b−a)⌉ bits to represent the number. We set \lceil \lg(b - a) \rceil⌈lg(b−a)⌉ as kk. Basically, we need to call \text{RANDOM}(0, 1)RANDOM(0,1) kk times. If the number represented by binary is bigger than b - ab−a, it's not valid number and we give it another try, otherwise we return that number.

RANDOM(a, b)
    range = b - a
    bits = ceil(log(2, range))
    result = 0
    for i = 0 to bits - 1
        r = RANDOM(0, 1)
        result = result + r << i
    if result > range
        return RANDOM(a, b)
    else return a + result

The expectation of times of calling procedure \text{RANDOM}(a, b)RANDOM(a,b) is \frac{2^k}{b - a}b−a2k​. \text{RANDOM}(0, 1)RANDOM(0,1) will be called kk times in that procedure.

The expected running time is \Theta(\frac{2^k}{b - a} \cdot k)Θ(b−a2k​⋅k), kk is \lceil \lg(b - a) \rceil⌈lg(b−a)⌉. Considering 2^k2k is less than 2 \cdot (b - a)2⋅(b−a), so the running time is O(k)O(k).

此处期望运行时间怎么算？