## Tag Archives: dynamic knapsack

## C++ || Knapsack Problem Using Dynamic Programming

The following is another homework assignment which was presented in an Algorithm Engineering class. Using a custom timer class, the following is a program which reduces the problem of selecting which Debian Linux packages to include on installation media, to the classical knapsack problem.

The program demonstrated on this page extends the previously implemented Exhaustive Search algorithm, this time using Dynamic Programming to find a solution to this problem.

**==== 1. THE PROBLEM ====**

The Exhaustive Search algorithm ran in O(2n) time and was very slow.

The goal of this program is to implement a dynamic programming algorithm which runs in O(n2 • W) time, returning the optimal set of items/packages contained in the input file, which is fast enough to produce solutions to inputs of realistic size.

This program solves the following problem:

**Input**: A list of Debian packages, each with a size (in kilobytes) and number of votes, both of which are integers; and an integer `W`

representing the capacity of the install media (in kilobytes).

**Output**: The set of packages with total size `≤ W`

, such that the total number of package votes is maximized.

In terms of the knapsack problem, the Debian package * votes* correspond to knapsack

*, and our Debian package*

**values***correspond to knapsack*

**sizes***.*

**weights****==== 2. THE INPUT ====**

The following file (knapsack_packages.txt) contains the * name*,

*, and*

**size***for approximately 36,000 binary Debian packages that are currently available on the amd64 platform, and have popcon information.*

**votes**The file obeys the following format:

`[ number of packages ]`

[ name 0 ] [ space ] [ votes 0 ] [ space ] [ size 0 ] [ newline ]

[ name 1 ] [ space ] [ votes 1 ] [ space ] [ size 1 ] [ newline ]

[ name 2 ] [ space ] [ votes 2 ] [ space ] [ size 2 ] [ newline ]

...

The packages are arranged within the file in sorted order, from the largest number of votes to the smallest.

**==== 3. FLOW OF CONTROL ====**

A test harness program is created which executes the above function and measures the elapsed time of the code corresponding to the algorithm in question. The test program will perform the following steps:

1.Load the "knapsack_packages.txt" file into memory.

2.Input size "n" and weight "W" as variables.

3.Execute the dynamic knapsack algorithm. The output of this algorithm should be the best combination sequence of packages.

4.Print the first 20 packages from the solution. A solution is the best sequence of packages, displaying theof the package, thename, and thevotesof each package, as well as the total votes and total size of the entire solution.size

A knapsack problem instances is created of varying input sizes “n” by using the first “n” entries in the file knapsack_packages.txt. In other words, to create a problem instance with `n = 100`

, only use the first 100 packages listed in the file as input.

**==== 4. TEST HARNESS ====**

**Note**: This program uses two external header files (Timer.h and Project2.h).

• Code for the Timer class (**Timer.h**) can be found here.

• Code for “**Project2.h**” can be found here.

• “**Project5.h**” is listed below.

```
```
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108
// =============================================================================// Author: K Perkins// Date: Nov 3, 2013// Taken From: http://programmingnotes.org/// File: main.cpp// Description: This is the test harness program which runs the code and // measures the elapsed time of the code corresponding to the algorithm // in question. //// This test program will reduce the problem of selecting which Debian// Linux packages to include on installation media, using the knapsack// problem. The goal is to implement a dynamic programming algorithm// for the knapsack problem, returning an optimal set of items/packages// which is fast enough to produce solutions to inputs of realistic size.// =============================================================================#include <iostream>#include <cstdlib>#include <cassert>#include <fstream>#include <vector>#include "Timer.h"#include "Project2.h"#include "Project5.h"using namespace std; // the maximum weightconst int MAX_WEIGHT = 65536; // the number of packages to examineconst int NUM_PACKAGES = 36149; int main(){ // declare variables Timer timer; Project5 proj5; Project2 proj2; PackageStats access; ifstream infile; vector<PackageStats> packages; vector<PackageStats> bestCombo; vector<int> knapResult; int* packageWeight = new int[NUM_PACKAGES]; int* packageVotes = new int[NUM_PACKAGES]; int totPackages = 0; // open file & make sure it exists infile.open("knapsack_packages.txt"); if(infile.fail()) { cout<<"nCant find file!n"; exit(1); } // get the total number of packages from the file infile >> totPackages; // make sure there are enough packages in the file assert(NUM_PACKAGES <= totPackages); // get the remaining info from the file // std::string int int for(int x=0;(infile >> access.name >> access.votes >> access.size) && x < NUM_PACKAGES; ++x) { packages.push_back(access); packageWeight[x] = access.size; packageVotes[x] = access.votes; } infile.close(); // display stats cerr<<"n = "<<NUM_PACKAGES<<", W = "<<MAX_WEIGHT <<"nn-- Dynamic Search Solution --n"; // start the timer timer.Start(); // return a vector containing the array indexes // of the best knapsack package solution knapResult = proj5.DynamicKnapsack(packageWeight, packageVotes, NUM_PACKAGES, MAX_WEIGHT); // stop the timer timer.Stop(); // using the data found from above, return // the packages that reside in those array indexes bestCombo = proj5.ReturnBest(packages, knapResult); // display info to the screen cout<<"n- Number of packages generated in this set: "<<bestCombo.size() <<"nn- First 20 packages..nn"; // display the best solution packages proj2.Display(bestCombo, 20); // display the size and total votes cout<<"nTotal Size = "<<proj2.TotalSize(bestCombo)<<" -- Total Votes = " <<proj2.TotalVotes(bestCombo)<<endl; // display the elapsed time cout<<endl<<"It took "<<timer.Elapsed()*1000 <<" clicks ("<<timer.Elapsed()<<" seconds)"<<endl; delete[] packageWeight; delete[] packageVotes; return 0;}// http://programmingnotes.org/

**==== 5. THE ALGORITHMS – “include Project5.h” ====**

```
```
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139
// =============================================================================// Author: K Perkins// Date: Nov 3, 2013// Taken From: http://programmingnotes.org/// File: Project5.h// Description: This is a simple class which holds the functions for // project 5// =============================================================================#ifndef PROJECT5_H#define PROJECT5_H #include <vector>#include <algorithm>#include "Project2.h"using std::vector; struct PackageSet{ int index; int weight; vector<int> appearances;}; class Project5{public: Project5(){} vector<int> DynamicKnapsack(int packageWeight[], int packageVotes[], int numPackages, int maxWeight) { // declare variables vector<int> best; vector<int> temp; vector<PackageSet> pset; int** T = new int*[numPackages+1]; int deleted = 0; int currWeight = maxWeight; bool dontCheck = false; for(int x=0; x <= numPackages; ++x) { T[x] = new int[maxWeight+1]; PackageSet access; dontCheck = false; temp.clear(); for(int y=0; y <= maxWeight; ++y) { if(x==0 || y==0) { T[x][y] = 0; } else if(packageWeight[x-1] <= y) { T[x][y] = std::max(T[x-1][y-packageWeight[x-1]] + packageVotes[x-1], T[x-1][y]); if(T[x][y] == (T[x-1][y-packageWeight[x-1]] + packageVotes[x-1])) { // if we find a valid packet, place it into // the temp vector for storage if(!dontCheck && !(std::find(temp.begin(), temp.end(), packageWeight[x-1]) != temp.end())) { temp.push_back(packageWeight[x-1]); } // find all of the weight instances where // this packet is "valid" access.appearances.push_back(y); dontCheck = true; } } else { T[x][y] = T[x-1][y]; } }// end for loop // gather info about the packet we just found if((std::find(temp.begin(), temp.end(), packageWeight[x-1]) != temp.end())) { access.index = x-1; access.weight = packageWeight[x-1]; pset.push_back(access); } // memory management, used to delete // array indexes thats no longer in use if(x > 1) { delete[] T[deleted++]; } }// end for loop delete[] T; // obtain the best possible knapsack solutuion, and save // their array indexes into a vector, starting from the end // NOTE: this places the knapsack solution in opposite (reverse) order for(int x = pset.size()-1; x >= 0; --x) { if(IsSolution(pset, x, currWeight)) { best.push_back(pset.at(x).index); currWeight -= pset.at(x).weight; } pset.pop_back(); } // reverse the vector back into ascending order std::reverse(best.begin(), best.end()); return best; }// end of DynamicKnapsack bool IsSolution(vector<PackageSet>& pset, int index, int currWeight) { return std::find(pset.at(index).appearances.begin(), pset.at(index).appearances.end(), currWeight) != pset.at(index).appearances.end(); }// end of IsSolution vector<PackageStats> ReturnBest(vector<PackageStats>& packages, vector<int>& knapResult) { vector<PackageStats> best; for(unsigned x=0; x < knapResult.size(); ++x) { best.push_back(packages.at(knapResult.at(x))); } return best; }// end of ReturnBest ~Project5(){}};#endif // http://programmingnotes.org/

**QUICK NOTES**:

The highlighted lines are sections of interest to look out for.

The code is heavily commented, so no further insight is necessary. If you have any questions, feel free to leave a comment below.

Note: Dont forget to include the input file!

The following is sample output:

=== RUN #1 ===

n = 36149, W = 65536

-- Dynamic Search Solution --

- Number of packages generated in this set: 764

- First 20 packages..

debianutils 90 128329

libgcc1 46 128327

debconf 168 121503

grep 595 121426

gzip 142 121346

sed 261 121220

findutils 805 121173

lsb-base 26 121121

sysv-rc 80 121092

sysvinit-utils 116 121078

base-files 65 121072

initscripts 93 121037

util-linux 846 121022

mount 272 120961

libselinux1 109 120918

cron 120 120872

base-passwd 49 120856

tzdata 488 120813

logrotate 64 120745

popularity-contest 67 120729Total Size = 65536 -- Total Votes = 25250749

It took 33500 clicks (33.50 seconds)

=== RUN #2 ===n = 36149, W = 800000

-- Dynamic Search Solution --

- Number of packages generated in this set: 4406

- First 20 packages..

debianutils 90 128329

libgcc1 46 128327

dpkg 2672 128294

perl-base 1969 127369

debconf 168 121503

grep 595 121426

gzip 142 121346

login 980 121332

coreutils 6505 121240

bash 1673 121229

sed 261 121220

findutils 805 121173

lsb-base 26 121121

sysv-rc 80 121092

sysvinit-utils 116 121078

base-files 65 121072

initscripts 93 121037

util-linux 846 121022

mount 272 120961

libselinux1 109 120918Total Size = 800000 -- Total Votes = 41588693

`It took 608760 clicks (608.76 seconds)`