June 02, 2020
The Knapsack Problem is a problem in combinatorial optimization which means one has to find optimal object from a set of finite objects. And it is another one of the most popular problem in computer science.
Letās first understand what is the ask in this problem? What do we need to find out?
We will be given a bunch of items with certain weights w and values v along with maximum weight W a Knapsack can hold. Our task is to choose the items whose combined weights is less than or equal to W with the maximum value. Any item can either be selected or not selected (0 or 1).
Waitā¦ what? š confused? Youāre supposed to be. But donāt worry, we can understand this together in simpler terms. Whatās better than a real world example to understand something.
Suppose your neighborhood store is giving an offer where they will give you a shopping cart and will ask you to go inside the store to fill the cart with any item (with a certain value) of your choice. You can choose as many items as you can š. But thereās a catchā¦ the only condition is that the items you choose cannot overfill the cart. Also, you cannot pick an item more than once.
What will you do? Of course, you will try to choose the values so that your selection will have the maximum values keeping in mind that your cart doesnāt get overfilled. Does this ring a bell š? Yes, this is exactly we are trying to accomplish in Knapsack problem.
You are given a fixed weight W (shopping cartās capacity) and items with a certain weight w and value v. Your task is to maximize the combined value keeping total weight of items <= W. The 0/1 comes from the fact that we can either select an item or not select an item. We cannot take a part of it.
Letās take an example to understand more about this problem. You are given the following -
value[] = {60, 100, 120}
.weight[] = {10, 20, 30}
.W = 50
.Thus, we want to make sure we choose items whose combined weights is less than or equal to 50 as well as they are of combined maximum value.
Now, a naive solution is to consider all subsets of items and calculate their combined weights and values. Now, choose only subsets whose combined weight is less than 50. Among those subsets, choose the one with maximum value.
Case 1
Select item with weight 10 alone V = 60
. Since 10 < 50
, thereās room for more.
Case 2
Select item with weight 20 alone V = 100
. Since 20 < 50
, thereās room for more.
Case 3
Select item with weight 30 alone V = 120
. Since 30 < 50
, thereās room for more.
Case 4
Select items with weight 10 and 20 V = 60 + 100 = 160
. Since 10 + 20 = 30 < 50
, thereās room for more.
Case 5
Select items with weight 10 and 30 V = 60 + 120 = 180
. Since 10 + 30 = 40 < 50
, thereās room for more.
Case 6
Select items with weight 20 and 30 V = 100 + 120 = 220
. Since 20 + 30 = 50 == 50
, this can be the solution.
Case 7
Select all items V = 60 + 100 + 120 = 280
. Since 10 + 20 + 30 = 60 > 50
, this cannot be the solution.
From above, we can easily deduce that items with weight 20 and 30 are the correct combination to choose.
But come on, do we really want to do this š? Donāt we have a better solution? In fact, we do! The answer is good old dynamic programming.
Let us randomly select an item from the list of items. For e.g., let me choose item with weight 20. I put this in the Knapsack, now, what is the capacity left of the Knapsack? The answer is, 50(total weight W) - 20(weight of the selected item) = 30. At this point, we only have to solve the problem for W = 30
. If we donāt select 20, then we will still have W = 50
and remaining list of items. Do you see the pattern? Yes, at each step the problem is reduced to a smaller problem.
We need to define a function that will give us the maximum value. Let this function be defined as -
int getMaximumValue(W, n, w[], v[])
where,
W => Maximum capacity of the Knapsack
n => Number of items to choose from
w[] => Array of weights of each item
v[] => Array of values of each item
Now, we can solve this problem by using the following steps -
Either we choose the i
th item with weight wi
and value vi
. In this case, number of items remaining to choose from will be n-1
. The function call will look like -
int getMaximumValue(W-wi, n-1, w[], v[])
Or, we donāt choose the item. In this case, the function call will look like -
int getMaximumValue(W, n-1, w[], v[])
Since we have to maximize the value, we will have to consider the maximum of the output of the above two conditions as -
max(getMaximumValue(W-wi, n-1, w[], v[]) + vi, getMaximumValue(W, n-1, w[], v[]))
In the first argument, we added vi because we included the item in the Knapsack while we didnāt in the second case. Therefore, to solve this problem, we only have to call this function recursively until either W or n becomes zero.
Have you noticed something? Yes, we have identified just encountered Optimal Substructure property of DP. Good job! š
But, if you percolate the function calls, you will find that we are solving a subproblem more than once. This is not ideal, right?
What have we encountered now? Yes, Overlapping Subproblems.
To tackle above, we will have to generate a lookup table of already calculated results. This will make our solution faster.
So much for the theory, eh? Now, letās dive into the interesting part. Coding!!!
public class Knapsack {
public static int getMaximumValue(int W, int n, int[] weights, int[] values) {
// Create the lookup table
int[][] lookup = new int[n + 1][W + 1];
// Solve for each case
for (int i = 0; i <= n; i++) {
for (int j = 0; j <= W; j++) {
// Base condition - no item, no capacity
if (i == 0 || j == 0) {
lookup[i][j] = 0;
}
// If we include the i-th item
else if (weights[i - 1] <= j) {
lookup[i][j] = Math.max(values[i - 1] + lookup[i - 1][j - weights[i - 1]], lookup[i - 1][j]);
}
// If we have out of the capacity
else {
lookup[i][j] = lookup[i - 1][j];
}
}
}
return lookup[n][W];
}
public static void main(String[] args) {
int[] weights = new int[]{10, 20, 30};
int[] values = new int[]{60, 100, 120};
int W = 50;
int n = weights.length;
System.out.println("Maximum value: " + getMaximumValue(W, n, weights, values));
}
}
def get_maximum_value(W, n, weights, values):
# Lookup table for memoization
lookup = [[0 for i in range(W + 1)] for j in range(n + 1)]
# Loop for items
for i in range(n + 1):
for j in range(W + 1):
# Base condition - no item, no capacity
if i == 0 or j == 0:
lookup[i][j] = 0
# For the i-th item
elif weights[i - 1] <= j:
lookup[i][j] = max(values[i - 1] + lookup[i - 1][j - weights[i - 1]], lookup[i - 1][j])
# Take the values from the previous row
else:
lookup[i][j] = lookup[i - 1][j]
return lookup[n][W]
weights = [10, 20, 30]
values = [60, 100, 120]
W = 50
n = len(weights)
print("Maximum value: ", get_maximum_value(W, n, weights, values))
import kotlin.math.max
fun getMaximumValue(W: Int, n: Int, weights: IntArray, values: IntArray): Int {
// Lookup table
val lookup = Array(n + 1) { i -> Array(W + 1) { j -> 0 } }
// Loop for each case
for (i in 0..n) {
for (j in 0..W) {
// Base condition - no item, no capacity
if (i == 0 || j == 0) {
lookup[i][j] = 0
}
// For the i-th element
else if (weights[i - 1] <= j) {
lookup[i][j] = max(values[i - 1] + lookup[i - 1][j - weights[i - 1]], lookup[i - 1][j])
}
// For all other cases take value from the previous row
else {
lookup[i][j] = lookup[i - 1][j]
}
}
}
return lookup[n][W]
}
fun main() {
val weights = intArrayOf(10, 20, 30)
val values = intArrayOf(60, 100, 120)
val W = 50
val n = weights.size
println("Maximum sum: " + getMaximumValue(W, n, weights, values))
}
function getMaximumValue(W, n, weights, values) {
// Lookup table
let lookup = Array.from(Array(n + 1), () => new Array(W + 1))
// Check for each case
for (let i = 0; i <= n; i++) {
for (let j = 0; j <= W; j++) {
// Base condition - no item, no capacity
if (i == 0 || j == 0) {
lookup[i][j] = 0;
}
// If we consider the i-th element
else if (weights[i - 1] <= j) {
lookup[i][j] = Math.max(values[i - 1] + lookup[i - 1][j - weights[i - 1]], lookup[i - 1][j]);
}
// For all other cases, get the values from previous row
else {
lookup[i][j] = lookup[i - 1][j];
}
}
}
return lookup[n][W];
}
function create2DArray(rows, columns) {
let lookup = [];
}
const weights = [10, 20, 30];
const values = [60, 100, 120];
const W = 50;
const n = weights.length;
console.log("Maximum value: " + getMaximumValue(W, n, weights, values));
The time and space complexity of this approach is O(n)(W)
.
So, in this post, we learned about yet another problem that sounds tricky but actually pretty simple. I hope you enjoyed this post.
You can find the code on my GitHub repository. If you like it, please share your thoughts in the comment.
See ya š.