[英]Dynamic Programming: Convert bottom-up approach to top-down on Problem 31 on ProjectEuler
For the sake of learning, I want to solve [problem 31] ( https://projecteuler.net/problem=31 ) on Project Euler using top-down DP.为了学习,我想使用自上而下的 DP 解决 Project Euler 上的 [问题 31] ( https://projecteuler.net/problem=31 )。 I have already solved it using both brute force and bottom-up Dynamic Programming.
我已经使用蛮力和自下而上的动态编程解决了它。
Problem summary: Number of unique combinations (order doesn't matter) in which sums to £2, having the coins 1p, 2p, 5p, 10p, 20p, 50p, £1, £2.问题总结:总和为 2 英镑的独特组合的数量(顺序无关紧要),硬币为 1p、2p、5p、10p、20p、50p、1 英镑、2 英镑。
I have done several attempts, but my result does take order into account.我做了几次尝试,但我的结果确实考虑到了顺序。 I tried to exclude this by including an "upperLimit" variable, whose job is to never include a more valuable coin (eliminating possibilities like 5,1,2,1, instead of 5,2,1,1)
我试图通过包含一个“upperLimit”变量来排除这一点,它的工作是永远不会包含一个更有价值的硬币(消除像 5,1,2,1 这样的可能性,而不是 5,2,1,1)
However, it still doesn't yield correct results.但是,它仍然没有产生正确的结果。 This is my code, and below the TOP DOWN I've included the bottom-up approach which works.
这是我的代码,在 TOP DOWN 下面我包含了有效的自下而上方法。
Lets simplify for 5p instead of 200p, with the coins 1,2,5.让我们简化为 5 便士而不是 200 便士,硬币为 1、2、5。 This should yield 4 combinations: 5, 221, 2111, 11111.
这应该产生 4 种组合:5、221、2111、11111。
My code with top-down outputs 5 combinations, while the bottom up correctly outputs 4.我的自上而下输出 5 种组合的代码,而自下而上正确输出 4 种组合。
TOP-DOWN APPROACH (doesn't work yet)自上而下的方法(还没有用)
combos = [0 for x in range(0,201)]
def combinations(currentValue, upperLimit = 200):
#Reset counter
numbOfCombos = 0
#If we reach leaf
if currentValue == 0:
return 1
#If the value is already known, return it
elif combos[currentValue] != 0:
return combos[currentValue]
#Else recurse through the tree
else:
if currentValue >= 200:
numbOfCombos = numbOfCombos + combinations(currentValue-200, 200)
if currentValue >= 100 and upperLimit >= 100:
numbOfCombos = numbOfCombos + combinations(currentValue-100, 100)
if currentValue >= 50 and upperLimit >= 50:
numbOfCombos = numbOfCombos + combinations(currentValue-50, 50)
if currentValue >= 20 and upperLimit >= 20:
numbOfCombos = numbOfCombos + combinations(currentValue-20, 20)
if currentValue >= 10 and upperLimit >= 10:
numbOfCombos = numbOfCombos + combinations(currentValue-10, 10)
if currentValue >= 5 and upperLimit >= 5:
numbOfCombos = numbOfCombos + combinations(currentValue-5, 5)
if currentValue >= 2 and upperLimit >= 2:
numbOfCombos = numbOfCombos + combinations(currentValue-2, 2)
if currentValue >= 1 and upperLimit >= 1:
numbOfCombos = numbOfCombos + combinations(currentValue-1, 1)
combos[currentValue] = numbOfCombos
return combos[currentValue]
print(combinations(5,))
BOTTOM-UP APPROACH (works)自下而上的方法(有效)
targetValue = 200;
coins = [1, 2, 5, 10, 20, 50, 100, 200];
combinations = [0 for x in range(0,targetValue+1)];
combinations[0] = 1;
for i in range(0, len(coins)):
for j in range(coins[i], targetValue+1):
combinations[j] = combinations[j] + combinations[j - coins[i]];
print(combinations);
Output
Any tips/advice or complete solutions are greatly appreciated.非常感谢任何提示/建议或完整的解决方案。 I know that the bottom-up solution is probably the most efficient and most beautiful, but for the sake of learning thought processes I'd like to solve it using TOP-DOWN.
我知道自下而上的解决方案可能是最有效和最漂亮的,但为了学习思维过程,我想使用自上而下解决它。
Thanks!谢谢!
It's counting 1,1,2,1 because when it's counting solutions with 2,*, it memoizes 2 solutions for 3p, then counts 2 solutions for 1,1,*.它正在计算 1,1,2,1,因为当它计算 2,* 的解决方案时,它会为 3p 记住 2 个解决方案,然后为 1,1,* 计算 2 个解决方案。
The fix is to include upperLimit
in the memoization.解决方法是在记忆中包含
upperLimit
。 (I also added a missing test upperLimit >= 200
.) See below. (我还添加了一个缺失的测试
upperLimit >= 200
。)见下文。
combos = {}
def combinations(currentValue, upperLimit=200):
# Reset counter
numbOfCombos = 0
# If we reach leaf
if currentValue == 0:
return 1
# If the value is already known, return it
elif (currentValue, upperLimit) in combos:
return combos[currentValue, upperLimit]
# Else recurse through the tree
else:
if currentValue >= 200 and upperLimit >= 200:
numbOfCombos = numbOfCombos + combinations(currentValue - 200, 200)
if currentValue >= 100 and upperLimit >= 100:
numbOfCombos = numbOfCombos + combinations(currentValue - 100, 100)
if currentValue >= 50 and upperLimit >= 50:
numbOfCombos = numbOfCombos + combinations(currentValue - 50, 50)
if currentValue >= 20 and upperLimit >= 20:
numbOfCombos = numbOfCombos + combinations(currentValue - 20, 20)
if currentValue >= 10 and upperLimit >= 10:
numbOfCombos = numbOfCombos + combinations(currentValue - 10, 10)
if currentValue >= 5 and upperLimit >= 5:
numbOfCombos = numbOfCombos + combinations(currentValue - 5, 5)
if currentValue >= 2 and upperLimit >= 2:
numbOfCombos = numbOfCombos + combinations(currentValue - 2, 2)
if currentValue >= 1 and upperLimit >= 1:
numbOfCombos = numbOfCombos + combinations(currentValue - 1, 1)
combos[currentValue, upperLimit] = numbOfCombos
return numbOfCombos
print(combinations(5))
Posted top down approach can be simplified (ie less conditionals) and generalized to any set of coins using Coin Change |使用Coin Change | Posted top down 方法可以简化(即更少的条件)并推广到任何一组硬币。 DP-7
DP-7
For a set of coin S, we split the solution space into two sets:对于一组硬币 S,我们将解空间分成两组:
Code代码
def count(S, target):
'''
Returns the count of ways we can sum
S[0...m-1] coins to get sum target
'''
def recur(m, target):
# Helper function to enable Memoization
# If target is 0 then there is 1
# solution (do not include any coin)
if (target == 0):
return 1
# target < 0 -> no solution exists
if (target < 0):
return 0
# no coins and target > 0 -> no solution
if (m <=0 and target > 0):
return 0
# Check if coin combination and target amount have been solved before
if (m, target) in lookup:
return lookup[(m, target)]
# sum of two solutions:
# 1. does not use m coin i.e. recur (m-1, target)
# 2. uses at least one of m-th coin i.e. recur(m, target - S[m-1])
# including S[m-1] (ii) excluding S[m-1]
lookup[(m, target)] = recur(m - 1, target ) + recur(m, target - S[m-1])
return lookup[(m, target)]
lookup = {}
return recur(len(S), target)
Test测试
arr = [1, 2, 5, 10, 20, 50, 100, 200]
print(count(arr, 200))
# Output: 73682
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.