jrr/clojure_knapsack

doll-smuggler

Solution for the "Doll Smuggler" knapsack problem at https://github.com/micahalles/doll-smuggler

Notes

I started with brute force because i) I love combinatorial explosion and ii) it'd be useful for developing (small) testcases. With the given testcase, the exhaustive solver runs for about 2.5 minutes before java.lang.OutOfMemoryError on my system.

The first heuristic approach that comes to mind is to sort the items by descending efficiency, and start loading the knapsack from the top of the list. (i.e., favor (value/weight).) I can easily contrive a counterexample, though, so I'll have to do better. (see break_heuristics.input)

To build a correct solution that runs in reasonable time, I turned to Wikipedia. I studied the 0/1 dynamic programming approach (not the pseudocode, but rather the recursive definitions above it) and executed it a couple times on paper to wrap my head around it. Finally, with no small amount of println debugging, I managed to implement it in Clojure.

Usage

Run it like this:

$ lein run testcases/given.input

Or, to compare results for different approaches (brute force, heuristics, recursive):

$ lein run -v testcases/break_heuristics.input
max_weight:    200
total weight:  551
total value:   345

by weight:          $65 (151)
by value:           $100 (200)
by efficiency:      $130 (171)
exhaustively:       $180 (200)
recursive solution: $180 (200)

packed dolls:

name weight value
doc    100    90
seth   100    90

To run all the tests:

$ lein test

lein test doll-smuggler.core-test

Ran 9 tests containing 9 assertions.
0 failures, 0 errors.