TopCoder Problem Statement - GoodSubset

Determine all combinations of an array that when multiplied together equal a given value.

001: import java.util.ArrayList;
002: import java.util.Collection;
003: import java.util.List;
004:
005: public class GoodSubset {
006:
007: /*
008: * Stores a list of Integers along with their product. Avoids having
009: * to muliply the Integers together multiple times to get their product.
010: * This class was added after the solution was working in an attempt to
011: * improve performace. A lot of time was being spent re-calculating
012: * the products of these lists.
013: * Note: I don't think it helped much...
014: */
015: class ListAndProduct {
016: List ints;
017: int product;
018:
019: ListAndProduct(List ints, int product) {
020: this.ints = ints;
021: this.product = product;
022: }
023: }
024:
025: // Reset count each time it reaches this value.
026: public static final int MOD_VALUE = 1_000_000_007;
027:
028:
029: /*
030: * The method called by the tests. Obtains a Collection of ListAndProduct
031: * objects, and then counts the number of items whose product equals
032: * goodValue.
033: */
034: public int numberOfSubsets(int goodValue, int[] d) {
035:
036: if (d.length == 0) return 0;
037:
038: Collection possibleGoodSubsets =
039: getPossibleGoodSubsets(goodValue, d, 0);
040:
041: int count = 0;
042:
043: for (ListAndProduct lap : possibleGoodSubsets) {
044:
045: if (goodValue == lap.product) {
046: count++;
047: if (count == MOD_VALUE) count = 0;
048: }
049:
050: }
051:
052: return count;
053:
054: }
055:
056: /*
057: * Returns a Collection of ListAndProduct objects. The product of of
058: * each item will be a factor of goodValue.
059: */
060: private Collection getPossibleGoodSubsets(
061: int goodValue, int[] d, int index) {
062:
063: Collection possibleGoodSubSets;
064:
065: /*
066: * If this is the last element in d[] the create the return list,
067: * otherwise, obtain it by recursively calling getPossibleSubsets using
068: * the next index.
069: */
070: if (index == (d.length - 1)) {
071: possibleGoodSubSets = new ArrayList<>();
072:
073: } else {
074: possibleGoodSubSets = getPossibleGoodSubsets(
075: goodValue, d, index + 1);
076: }
077:
078: // Only interested in numbers that are factors of goodValue.
079: if (goodValue % d[index] == 0) {
080:
081: /*
082: * Contains new items to be added to possibleGoodSubSets. Don't
083: * want to modify that list until the for loop is done though.
084: */
085: List toAdd = new ArrayList<>();
086:
087: for (ListAndProduct lap : possibleGoodSubSets) {
088:
089: /*
090: * For each of the items currently in the possibleGoodSubSets
091: * Collection, does multiplying that product by d[index] also
092: * result in a factor of goodValue? If so, note it so it gets
093: * added to the Collection.
094: */
095: long product = (long) d[index] * (long) lap.product;
096:
097: if ((goodValue % product) == 0) {
098:
099: /*
100: * Create and store a new ListAndProduct object using the
101: * previous list of integers, and the current index. Store
102: * the list along with its product.
103: */
104: List newInts = new ArrayList<>();
105: for (int i : lap.ints) {
106: newInts.add(i);
107: }
108: newInts.add(d[index]);
109: toAdd.add(new ListAndProduct(newInts, (int) product));
110: }
111:
112: }
113:
114: /*
115: * possibleGoodSubSets could not be modified within the above for
116: * loop. So now, we can add all the items from toAdd into it.
117: */
118: for (ListAndProduct lap : toAdd) {
119: possibleGoodSubSets.add(lap);
120: }
121:
122: // The current index is a factor of goodValue, so add that in too.
123: List l = new ArrayList<>();
124: l.add(d[index]);
125: possibleGoodSubSets.add(new ListAndProduct(l, d[index]));
126: }
127:
128: return possibleGoodSubSets;
129:
130: }
131: }

First, a disclaimer. The given code works, but fails one of the timing tests. The constraints are that the solution
must complete in under 2 seconds. The code above passes all the tests except test #32. It give the correct answer, but
takes about 25 seconds to complete. I've tried a few techniques to bring this time down without success. If you've got
any ideas, I'd like to hear them.

The solution uses a technique know as dynamic programming. Here, we'll solve for the last element in the array; and then
using what we've computed, work backward toward the beginning of the array. I'll explain how it works, by walking through
an example.

For test3, the input array is { 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 } and goodValue = 12.

The idea is to start at the right side of the array, and only consider elements or products that are factors of goodValue.
Items matching our criteria are stored in a Collection of Lists of Integers called possibleGoodSubsets. So starting at 12:
12 is a factor of 12, so add it to possibleGoodSubsets. 11, 10, 9, 8, and 7 are not factors of 12, so we skip past those.

6 is a factor of 12, so add that. However 6 * 12 = 72 which is not a factor, so we don't add 6,12 to the List.

5 is not a factor, so skip it. 4 is a factor, so we'll add it just like 6 was.

3 is a factor of 12. Also 4 is already in our list, and 3 * 4 = 12 is a factor, so we'll add not only 3; but also 3,4

The table below shows how possibleGoodSubsets grows as we work backward through the array.

Value |
possibleGoodSubsets |

12 |
{{12}} |

11 |
{{12}} |

10 |
{{12}} |

9 |
{{12}} |

8 |
{{12}} |

7 |
{{12}} |

6 |
{{12}, {6}} |

5 |
{{12}, {6}} |

4 |
{{12}, {6}, {4}} |

3 |
{{12}, {6}, {4}, {3,4}, {3}} |

2 |
{{12}, {6}, {4}, {3,4}, {3}, {6,2}, {3,2}, {2}} |

1 |
{{12}, {6}, {4}, {3,4}, {3}, {6,2}, {3,2}, {2}, {12,1}, {6,1}, {4,1}, {3,4,1}, {3,1}, {6,2,1}, {3,2,1}, {2,1}, {1}} |

The collection given at Value 1 is what will be returned to the numOfSubsets() method. It then looks for elements whose product
equals goodValue. In this case {{12}, {3,4}, {6,2}, {12,1}, {3,4,1}, {6,2,1}}

This works great so long as not too many elements of the input array are factors of goodValue. In test32 all 100 of the items
in the input array are factors, and many of those number, when multiplied together produce factors. So the size of possibleGoodSubsets
really explodes in that case.

One thing I noticed when profiling the code was this it was spending most of its time computing the products. To alieviate this,
I introduced the class ListAndProduct so that I could store a List of Integers and it's product together, and thus avoid
having to calculate the product of that list over and over again. Unfortunately, the effectiveness of that chage was minimal.

I think this is a good algorithm, and does a good job of illustrating techniques of dynamic programming. So, I'm not really
interested in re-writing it. But, if you have ideas that can help get this under the 2 second limit for the worst case scenarios,
I'd like to hear it.