与R中的大数据集匹配

Question

I received this problem from a professor of mine to solve in R. Here's what I came up with:

buttons <- c(16,23,61,7,7,7,13,13,13,19,19,21,27,56,56,73,77,87,11,37,41)
combos<- NULL

for(bb in 1:80000){
  random<-sample(buttons,5,replace=FALSE)
  #A*B+C-D+E
  combos[bb]<-(random[1])*(random[2])+(random[3])-(random[4])+(random[5])
}

solutions<-c(917,134,1569,1649,1431,1622,233,2094,1072,915,
  1922,2437,2714,2491,1886,2812,426,1673,94,2139,2569,496,2249,1553,1580)

solutions %in% combos

What I think the code is doing:

1. Sampling 5 of the buttons on the machine without replacement.
2. Plugging those 5 numbers in the A*B+C-D+E formula.
3. Spitting out a final answer. Repeat 80,000 times.
4. Last command should check all the outputs from the formula against the snack numbers in the machine and return a Boolean.

However, the Boolean comes back with 25 Trues when there should be 1 false. Where did I go wrong?

Answer 1

I don't know why you try to solve this with random sampling. You could never trust any answer other than that you can get all snacks. I'd use this:

buttons <- as.integer(buttons)
solutions <- as.integer(solutions)

#create all combinations of 5 buttons
combos <- t(combn(buttons, 5))
library(combinat)
#permute the combinations
tmp <- lapply(permn(1:5), function(i, solutions, combos) {
  #which solutions can be derived from the permuted combination?
  solutions[solutions %in% (combos[,i[1]] * 
                                combos[,i[2]] + 
                                combos[,i[3]] - 
                                combos[,i[4]] + 
                                combos[,i[5]])]
}, solutions = solutions, combos = combos)

#which solution can not be achieved?   
solutions[!(solutions %in% unlist(tmp))]

However, this doesn't give me a snack I can't get either. Maybe I'm misinterpreting the wording.

Answer 2

For starters remove the for loop. It is very inefficient:

random <- matrix(replicate(8e4, sample(buttons,5,replace=F)), ncol=5, byrow=TRUE)
combos <-random[,1]*random[,2]+random[,3]-random[,4]+random[,5]

When I expand to one million combos, it matches all TRUE each time. There may be some confusion I'm not seeing in the question. Or your professor is a very cruel person.

Update:

I changed the buttons vector as mentioned in the comments:

solutions[!solutions %in% combos]
[1] 2437

Thanks, I can move on with my regularly scheduled life. Sampling isn't the best way, but problem solved.