第一种思路是,每个骰子的点数从最小到最大,假设为1-6,那么所有的骰子从最小1开始,我们假设一种从左向右的排列,右边的最低,索引从最低开始,判断和的情况。
def setTo1(dices, start, end):
for i in range(start, end):
dices[i] = 1
def probability(n, s, dmax = 6, dmin = 1):
if s < n * dmin or s > n * dmax : return 0
dices = [1] * n
i = n - 1
total = 0
while i >= 0:
curSum = sum(dices)
if curSum == s:
print dices
total += 1
# find first one that can +1
for j in range(i, -1, -1):
if dices[j] < dmax and s - sum(dices[0:j+1]) >= n - j*dmin:
dices[j] += 1
setTo1(dices, j + 1, n)
i = n - 1
break
else:
i -= 1
elif curSum < s:
if dices[i] < dmax:
dices[i] += 1
i = n - 1
else:
i -= 1
print "total = {0}, prob = {1}%".format(total, total*100/dmax**n)
return total
若当前和小于s,则检验当前索引处的骰子能否增加1,若能,则增加,否则查看其前面的能否增加。若相等,那么我们统计信息后,要变化当前的情形,以便处理下一种情况,因为 索引是从低位开始到当前位的,所以我们从当前索引开始,向前找能继续增加的骰子,这里的判断标准是当前骰子的点数小于最小值,而且要保证其后的骰子的最小值为1,比如 1,4,1, s = 6, 当前索引指向4, 这里的4虽然小于最大点数6, 但若其再加一,第三个骰子就的为0,这不符合要求。若找到可以加一的骰子,那就将该骰子点数加1, 将其后的骰子都置为1,索引回到最后,开始重新加起。如:1,1,6,s = 8, 索引指向6, 修改后为1,2,1,索引指向最后的1,。若没有找到可以再增加的骰子,那么就结束。如6,1,1,s = 8。
其实若给定的n不大的话,我们可以设一个n位整数,从n个1开始,逐次加一,来判断各个位的和是否满足要求,直到达到最大值,n个6。
这个问题其实动态规划的特点很明显。
'''
@ state function: dp[i, j]: the total cases of sum = j, composed by i dices
@ state tranfor function: dp[i, j] = sum(dp[i - 1, j - k]) for k in [dmin, dmax]
@ dp[i, j] = 0, j > i * dmax or j < i * dmin
@ init condition: dp[1, k] = 1, for k in [dmin, dmax], dp[1, k] = 0, for other k
'''
def dp_probability(n, s, dmax = 6, dmin = 1):
if s < n * dmin or s > n * dmax :
return 0
dp1 = [0] * (n * dmax + 1)
#init dp[1, :]
for i in range(1, dmax + 1):
dp1[i] = 1
# i: the number of dices
for i in range(2, n + 1):
dp2 = [0] * (n * dmax + 1)
# j: range of i dices
for j in range(dmin * i, dmax * i + 1):
# k: range of new added dice
for k in range(dmin, dmax + 1):
if j > k :
dp2[j] += dp1[j - k]
print dp2
dp1 = dp2
print "total = {0}, prob = {1}%".format(dp2[s], dp2[s]*100/dmax**n)
return dp2[s]