Dynamic Programming and the

Knapsack Problem
Paul Dohmen
Roshnika Fernando
What is Dynamic Programming

Dynamic programming is a method of

solving complex problems by breaking
them down into sub-problems that can be
solved by working backwards from the last
stage.

Coined by Richard Bellman who described

dynamic programming as the way of
solving problems where you need to find
the best decisions one after another
Dynamic Programming

Dynamic programming is a simple method

of solving complex real world problems,
such as
Traveling salesman problem
Fibonacci sequence
Stagecoach problem
Knapsack problem

Numerous other applications exist

Steps to Dynamic Programming

Every problem is divided into stages

Each stage requires a decision
Decisions are made to determine the state of
the next stage
The solution procedure is to find an optimal
solution at each stage for every possible
state
This solution procedure often starts at the
last stage and works its way forward
Knapsack Problem

You are a mischievous child camping in the

woods. You would like to steal items from
other campers, but you can only carry so
much mass in your knapsack. You see seven
items worth stealing. Each
item has a certain mass and
monetary value. How do you
maximize your profit so you
can buy more video games
later?
Knapsack Problem

Value and mass of

each item is given
Maximize profit
Subject to mass
constraint of
knapsack: 15 kg

Being a smart kid,

you apply dynamic
programming
Smart Kids Write Equations

There are potentially 7 stages to this dynamic

programming problem.
When an item is placed in the knapsack, a decision,
F(m), is made to choose the next optimal item, i.
Start at the end of the problem, assuming that the
knapsack has an unknown mass, m.
Recurrence relation:
F(i,m) = max (F(i-1,m-mi) + fi, F(i-1, m))

Use recursion and iterate through the problem,

eventually coming up with the optimal solution.
Smart Kids Use MatLab

Now imagine you, the mischievous smart

thief kid, brought your laptop to the
camping trip. How would you solve the
equation? MATLAB!

Matlab Solution:
Profit: $21
Mass: 15kg
Not enough to buy a new video game
Solution
Sensitivity Analysis

Valuables are not very valuable when camping,

pick richer people to steal from.

Suddenly a valuable item of $15 that weighs

11kg is available, will the optimal solution
change?
Yes, the maximum profit becomes $23

Using MatLab, it is easy to change the variables

on the program for specific sensitivity analysis
Questions?
Sources

Petter. "MATLAB Central - File detail - 0-1 Knapsack." The

MathWorks - MATLAB and Simulink for Technical
Computing. Web. 03 Dec. 2009.
<http://www.mathworks.com/matlabcentral/fileexchang
e/22783-0-1-knapsack>.
"Dynamic Programming Algorithms." Web. 03 Dec. 2009.
<http://www.personal.kent.edu/~rmuhamma/Algorithm
s/MyAlgorithms/Dynamic/knapsackdyn.htm>.
"Dynamic programming ." Wikipedia, the free encyclopedia.
Web. 03 Dec. 2009.
<http://en.wikipedia.org/wiki/Dynamic_programming>.
"Dynamic Programming: Definition and additional resources
from BNET." Glossary of Business Dictionary Terms | A |
BNET. Web. 02 Dec. 2009.
<http://dictionary.bnet.com/definition/dynamic+progra
mming.html>.