Editing Dynamic programming (section)

== Examples: computer algorithms ==

=== Dijkstra's algorithm for the shortest path problem ===
From a dynamic programming point of view, [[Dijkstra's algorithm]] for the [[shortest path problem]]  is a successive approximation scheme that solves the dynamic programming functional equation for the shortest path problem by the '''Reaching''' method.<ref name=sniedovich_06>{{Citation | last = Sniedovich | first = M. | title = Dijkstra's algorithm revisited: the dynamic programming connexion | journal = Journal of Control and Cybernetics | volume = 35 | issue = 3 | pages = 599–620 | year = 2006 | url = http://matwbn.icm.edu.pl/ksiazki/cc/cc35/cc3536.pdf | postscript = .}} [http://www.ifors.ms.unimelb.edu.au/tutorial/dijkstra_new/index.html Online version of the paper with interactive computational modules.]</ref><ref name=denardo_03>{{Citation | last = Denardo | first = E.V. | title = Dynamic Programming: Models and Applications | publisher = [[Dover Publications]] | location = Mineola, NY | year = 2003 | isbn = 978-0-486-42810-9}}</ref><ref name=sniedovich_10>{{Citation | last = Sniedovich | first = M. | title = Dynamic Programming: Foundations and Principles | publisher = [[Taylor & Francis]] | year = 2010 | isbn = 978-0-8247-4099-3  }}</ref>

In fact, Dijkstra's explanation of the logic behind the algorithm,<ref>{{cite journal |last=Dijkstra |first=E. W.|author-link=Edsger W. Dijkstra|title=A note on two problems in connexion with graphs |journal=Numerische Mathematik |date=December 1959 |volume=1 |issue=1 |pages=269–271 |doi=10.1007/BF01386390}}</ref> namely
{{blockquote|
'''Problem 2.''' Find the path of minimum total length between two given nodes <math>P</math> and <math>Q</math>.

We use the fact that, if <math>R</math> is a node on the minimal path from <math>P</math> to <math>Q</math>, knowledge of the latter implies the knowledge of the minimal path from <math>P</math> to <math>R</math>.
}}

is a paraphrasing of [[Richard Bellman|Bellman's]] famous [[Principle of Optimality]] in the context of the [[shortest path problem]].

=== Fibonacci sequence ===
Using dynamic programming in the calculation of the ''n''th member of the [[Fibonacci sequence]] improves its performance greatly. Here is a naïve implementation, based directly on the mathematical definition:

 '''function''' fib(n)
     '''if''' n <= 1 '''return''' n
     '''return''' fib(n − 1) + fib(n − 2)

Notice that if we call, say, <code>fib(5)</code>, we produce a call tree that calls the function on the same value many different times:

# <code>fib(5)</code>
# <code>fib(4) + fib(3)</code>
# <code>(fib(3) + fib(2)) + (fib(2) + fib(1))</code>
# <code>((fib(2) + fib(1)) + (fib(1) + fib(0))) + ((fib(1) + fib(0)) + fib(1))</code>
# <code>(((fib(1) + fib(0)) + fib(1)) + (fib(1) + fib(0))) + ((fib(1) + fib(0)) + fib(1))</code>

In particular, <code>fib(2)</code> was calculated three times from scratch. In larger examples, many more values of <code>fib</code>, or ''subproblems'', are recalculated, leading to an exponential time algorithm.

Now, suppose we have a simple [[Associative array|map]] object, ''m'', which maps each value of <code>fib</code> that has already been calculated to its result, and we modify our function to use it and update it. The resulting function requires only [[Big-O notation|O]](''n'') time instead of exponential time (but requires [[Big-O notation|O]](''n'') space):

 '''var''' m := '''''map'''''(0 → 0, 1 → 1)
 '''function''' fib(n)
     '''if ''key''''' n '''is not in ''map''''' m 
         m[n] := fib(n − 1) + fib(n − 2)
     '''return''' m[n]

This technique of saving values that have already been calculated is called ''[[memoization]]''; <!-- Yes, memoization, not memorization. Not a typo. --> this is the top-down approach, since we first break the problem into subproblems and then calculate and store values.

In the '''bottom-up''' approach, we calculate the smaller values of <code>fib</code> first, then build larger values from them. This method also uses O(''n'') time since it contains a loop that repeats n − 1 times, but it only takes constant (O(1)) space, in contrast to the top-down approach which requires O(''n'') space to store the map.

 '''function''' fib(n)
     '''if''' n = 0
         '''return''' 0
     '''else'''
         '''var''' previousFib := 0, currentFib := 1
         '''repeat''' n − 1 '''times''' ''// loop is skipped if n = 1''
             '''var''' newFib := previousFib + currentFib
             previousFib := currentFib
             currentFib  := newFib
         '''return''' currentFib

In both examples, we only calculate <code>fib(2)</code> one time, and then use it to calculate both <code>fib(4)</code> and <code>fib(3)</code>, instead of computing it every time either of them is evaluated.

=== A type of balanced 0–1 matrix ===
{{unreferenced section|date=May 2013}}
Consider the problem of assigning values, either zero or one, to the positions of an {{math|<var>n</var> &times; <var>n</var>}} matrix, with {{math|<var>n</var>}} even, so that each row and each column contains exactly {{math|<var>n</var> / 2}} zeros and {{math|<var>n</var> / 2}} ones. We ask how many different assignments there are for a given <math>n</math>. For example, when {{math|<var>n</var> {{=}} 4}}, five possible solutions are

:<math>\begin{bmatrix}
0 & 1 & 0 & 1 \\
1 & 0 & 1 & 0 \\
0 & 1 & 0 & 1 \\
1 & 0 & 1 & 0
\end{bmatrix} \text{ and } \begin{bmatrix}
0 & 0 & 1 & 1 \\
0 & 0 & 1 & 1 \\
1 & 1 & 0 & 0 \\
1 & 1 & 0 & 0
\end{bmatrix} \text{ and } \begin{bmatrix}
1 & 1 & 0 & 0 \\
0 & 0 & 1 & 1 \\
1 & 1 & 0 & 0 \\
0 & 0 & 1 & 1
\end{bmatrix} \text{ and } \begin{bmatrix}
1 & 0 & 0 & 1 \\
0 & 1 & 1 & 0 \\
0 & 1 & 1 & 0 \\
1 & 0 & 0 & 1
\end{bmatrix} \text{ and } \begin{bmatrix}
1 & 1 & 0 & 0 \\
1 & 1 & 0 & 0 \\
0 & 0 & 1 & 1 \\
0 & 0 & 1 & 1
\end{bmatrix}.</math>

There are at least three possible approaches: [[Brute-force search|brute force]], [[backtracking]], and dynamic programming.

Brute force consists of checking all assignments of zeros and ones and counting those that have balanced rows and columns ({{math|<var>n</var> / 2}} zeros and {{math|<var>n</var> / 2}} ones). As there are <math>2^{n^2}</math> possible assignments and <math>\tbinom{n}{n/2}^n</math> sensible assignments, this strategy is not practical except maybe up to <math>n=6</math>.

Backtracking for this problem consists of choosing some order of the matrix elements and recursively placing ones or zeros, while checking that in every row and column the number of elements that have not been assigned plus the number of ones or zeros are both at least {{math|<var>n</var> / 2}}. While more sophisticated than brute force, this approach will visit every solution once, making it impractical for {{math|<var>n</var>}} larger than six, since the number of solutions is already 116,963,796,250 for {{math|<var>n</var>}}&nbsp;=&nbsp;8, as we shall see.

Dynamic programming makes it possible to count the number of solutions without visiting them all. Imagine backtracking values for the first row – what information would we require about the remaining rows, in order to be able to accurately count the solutions obtained for each first row value? We consider {{math|<var>k</var> &times; <var>n</var>}} boards, where {{math|1 &le; <var>k</var> &le; <var>n</var>}}, whose <math>k</math> rows contain <math>n/2</math> zeros and <math>n/2</math> ones. The function ''f'' to which [[memoization]] is applied maps vectors of ''n'' pairs of integers to the number of admissible boards (solutions). There is one pair for each column, and its two components indicate respectively the number of zeros and ones that have yet to be placed in that column. We seek the value of <math> f((n/2, n/2), (n/2, n/2), \ldots (n/2, n/2)) </math> (<math>n</math> arguments or one vector of <math>n</math> elements). The process of subproblem creation involves iterating over every one of  <math>\tbinom{n}{n/2}</math> possible assignments for the top row of the board, and going through every column, subtracting one from the appropriate element of the pair for that column, depending on whether the assignment for the top row contained a zero or a one at that position. If any one of the results is negative, then the assignment is invalid and does not contribute to the set of solutions (recursion stops). Otherwise, we have an assignment for the top row of the {{math|<var>k</var> &times; <var>n</var>}} board and recursively compute the number of solutions to the remaining {{math|(<var>k</var> &minus; 1) &times; <var>n</var>}} board, adding the numbers of solutions for every admissible assignment of the top row and returning the sum, which is being memoized. The base case is the trivial subproblem, which occurs for a {{math|1 &times; <var>n</var>}} board. The number of solutions for this board is either zero or one, depending on whether the vector is a permutation of {{math|<var>n</var> / 2}} <math>(0, 1)</math> and {{math|<var>n</var> / 2}} <math>(1, 0)</math> pairs or not.

For example, in the first two boards shown above the sequences of vectors would be
<pre>
((2, 2) (2, 2) (2, 2) (2, 2))       ((2, 2) (2, 2) (2, 2) (2, 2))     k = 4
  0      1      0      1              0      0      1      1

((1, 2) (2, 1) (1, 2) (2, 1))       ((1, 2) (1, 2) (2, 1) (2, 1))     k = 3
  1      0      1      0              0      0      1      1

((1, 1) (1, 1) (1, 1) (1, 1))       ((0, 2) (0, 2) (2, 0) (2, 0))     k = 2
  0      1      0      1              1      1      0      0

((0, 1) (1, 0) (0, 1) (1, 0))       ((0, 1) (0, 1) (1, 0) (1, 0))     k = 1
  1      0      1      0              1      1      0      0

((0, 0) (0, 0) (0, 0) (0, 0))       ((0, 0) (0, 0), (0, 0) (0, 0))
</pre>

The number of solutions {{OEIS|id=A058527}} is

:<math> 1,\, 2,\, 90,\, 297200,\, 116963796250,\, 6736218287430460752, \ldots </math>

Links to the MAPLE implementation of the dynamic programming approach may be found among the [[#External links|external links]].

=== Checkerboard ===
{{unreferenced section|date=May 2013}}
Consider a [[checkerboard]] with ''n'' × ''n'' squares and a cost function <code>c(i, j)</code> which returns a cost associated with square <code>(i,j)</code> (<code>''i''</code> being the row, <code>''j''</code> being the column). For instance (on a 5 × 5 checkerboard),

{| class="wikitable" style="text-align:center"
|-
! 5
| 6 || 7 || 4 || 7 || 8
|-
! 4
| 7 || 6 || 1 || 1 || 4
|-
! 3
| 3 || 5 || 7 || 8 || 2
|-
! 2
| – || 6 || 7 || 0 || –
|-
! 1
| – || – || '''5''' || – || –
|-
!width="15"| !! style="width:15px;"|1 !! style="width:15px;"|2 !! style="width:15px;"|3 !! style="width:15px;"|4 !! style="width:15px;"|5
|}
Thus <code>c(1, 3) = 5</code>

Let us say there was a checker that could start at any square on the first rank (i.e., row) and you wanted to know the shortest path (the sum of the minimum costs at each visited rank) to get to the last rank; assuming the checker could move only diagonally left forward, diagonally right forward, or straight forward. That is, a checker on <code>(1,3)</code> can move to <code>(2,2)</code>, <code>(2,3)</code> or <code>(2,4)</code>.
{| class="wikitable" style="text-align:center"
|-
! 5
|  ||  ||  ||  ||
|-
! 4
|  ||  ||  ||  ||
|-
! 3
|  ||  ||  ||  ||
|-
! 2
|  || x  || x || x ||
|-
! 1
|  ||  || o ||  ||
|-
!width="15"| !! style="width:15px;"|1 !! style="width:15px;"|2 !! style="width:15px;"|3 !! style="width:15px;"|4 !! style="width:15px;"|5
|}

This problem exhibits '''optimal substructure'''. That is, the solution to the entire problem relies on solutions to subproblems. Let us define a function <code>q(i, j)</code> as

:''q''(''i'', ''j'') = the minimum cost to reach square (''i'', ''j'').

Starting at rank <code>n</code> and descending to rank <code>1</code>, we compute the value of this function for all the squares at each successive rank. Picking the square that holds the minimum value at each rank gives us the shortest path between rank <code>n</code> and rank <code>1</code>.

The function <code>q(i, j)</code> is equal to the minimum cost to get to any of the three squares below it (since those are the only squares that can reach it) plus <code>c(i, j)</code>. For instance:

{| class="wikitable" style="text-align:center"
|-
! 5
|  ||  ||  ||  ||
|-
! 4
|  ||  || A  ||  ||
|-
! 3
|  || B || C || D ||
|-
! 2
|  ||  ||  ||  ||
|-
! 1
|  ||  ||  ||  ||
|-
!width="15"| !! style="width:15px;"|1 !! style="width:15px;"|2 !! style="width:15px;"|3 !! style="width:15px;"|4 !! style="width:15px;"|5
|}

: <math>q(A) = \min(q(B),q(C),q(D))+c(A) \, </math>

Now, let us define <code>q(i, j)</code> in somewhat more general terms:

: <math>q(i,j)=\begin{cases} \infty & j < 1 \text{ or }j > n \\ c(i, j) & i = 1 \\ \min(q(i-1, j-1), q(i-1, j), q(i-1, j+1)) + c(i,j) & \text{otherwise.}\end{cases}</math>

The first line of this equation deals with a board modeled as squares indexed on <code>1</code> at the lowest bound and <code>n</code> at the highest bound. The second line specifies what happens at the first rank; providing a base case. The third line, the recursion, is the important part. It represents the <code>A,B,C,D</code> terms in the example. From this definition we can derive straightforward recursive code for <code>q(i,&nbsp;j)</code>. In the following pseudocode, <code>n</code> is the size of the board, <code>c(i, j)</code> is the cost function, and <code>min()</code> returns the minimum of a number of values:

 '''function''' minCost(i, j)
     '''if''' j &lt; 1 '''or''' j > n
         '''return''' infinity
     '''else if''' i = 1
         '''return''' c(i, j)
     '''else'''
         '''return''' '''min'''( minCost(i-1, j-1), minCost(i-1, j), minCost(i-1, j+1) ) + c(i, j)

This function only computes the path cost, not the actual path. We discuss the actual path below. This, like the Fibonacci-numbers example, is horribly slow because it too exhibits the '''overlapping sub-problems''' attribute. That is, it recomputes the same path costs over and over. However, we can compute it much faster in a bottom-up fashion if we store path costs in a two-dimensional array <code>q[i, j]</code> rather than using a function. This avoids recomputation; all the values needed for array <code>q[i, j]</code> are computed ahead of time only once. Precomputed values for <code>(i,j)</code> are simply looked up whenever needed.

We also need to know what the actual shortest path is. To do this, we use another array <code>p[i, j]</code>; a ''predecessor array''. This array records the path to any square <code>s</code>. The predecessor of <code>s</code> is modeled as an offset relative to the index (in <code>q[i, j]</code>) of the precomputed path cost of <code>s</code>. To reconstruct the complete path, we lookup the predecessor of <code>s</code>, then the predecessor of that square, then the predecessor of that square, and so on recursively, until we reach the starting square. Consider the following pseudocode:

 '''function''' computeShortestPathArrays()
     '''for''' x '''from''' 1 '''to''' n
         q[1, x] := c(1, x)
     '''for''' y '''from''' 1 '''to''' n
         q[y, 0]     := infinity
         q[y, n + 1] := infinity
     '''for''' y '''from''' 2 '''to''' n
         '''for''' x '''from''' 1 '''to''' n
             m := min(q[y-1, x-1], q[y-1, x], q[y-1, x+1])
             q[y, x] := m + c(y, x)
             '''if''' m = q[y-1, x-1]
                 p[y, x] := -1
             '''else if''' m = q[y-1, x]
                 p[y, x] :=  0
             '''else'''
                 p[y, x] :=  1

Now the rest is a simple matter of finding the minimum and printing it.

 '''function''' computeShortestPath()
     computeShortestPathArrays()
     minIndex := 1
     min := q[n, 1]
     '''for''' i '''from''' 2 '''to''' n
         '''if''' q[n, i] &lt; min
             minIndex := i
             min := q[n, i]
     printPath(n, minIndex)

 '''function''' printPath(y, x)
     '''print'''(x)
     '''print'''("&lt;-")
     '''if''' y = 2
         '''print'''(x + p[y, x])
     '''else'''
         printPath(y-1, x + p[y, x])

=== Sequence alignment ===
In [[genetics]], [[sequence alignment]] is an important application where dynamic programming is essential.<ref name="Eddy"/>  Typically, the problem consists of transforming one sequence into another using edit operations that replace, insert, or remove an element.  Each operation has an associated cost, and the goal is to find the [[edit distance|sequence of edits with the lowest total cost]].

The problem can be stated naturally as a recursion, a sequence A is optimally edited into a sequence B by either:

# inserting the first character of B, and performing an optimal alignment of A and the tail of B
# deleting the first character of A, and performing the optimal alignment of the tail of A and B
# replacing the first character of A with the first character of B, and performing optimal alignments of the tails of A and B.

The partial alignments can be tabulated in a matrix, where cell (i,j) contains the cost of the optimal alignment of A[1..i] to B[1..j].  The cost in cell (i,j) can be calculated by adding the cost of the relevant operations to the cost of its neighboring cells, and selecting the optimum.

Different variants exist, see [[Smith–Waterman algorithm]] and [[Needleman–Wunsch algorithm]].

=== Tower of Hanoi puzzle ===
[[Image:Tower of Hanoi.jpeg|upright=1.2|thumb|A model set of the Towers of Hanoi (with 8 disks)]]
[[Image:Tower of Hanoi 4.gif|upright=1.2|thumb|An animated solution of the '''Tower of Hanoi''' puzzle for ''T(4,3)''.]]
The '''[[Tower of Hanoi]]''' or '''Towers of [[Hanoi]]''' is a [[mathematical game]] or [[puzzle]]. It consists of three rods, and a number of disks of different sizes which can slide onto any rod. The puzzle starts with the disks in a neat stack in ascending order of size on one rod, the smallest at the top, thus making a conical shape.

The objective of the puzzle is to move the entire stack to another rod, obeying the following rules:

* Only one disk may be moved at a time.
* Each move consists of taking the upper disk from one of the rods and sliding it onto another rod, on top of the other disks that may already be present on that rod.
* No disk may be placed on top of a smaller disk.

The dynamic programming solution consists of solving the [[Bellman equation|functional equation]]

: S(n,h,t) = S(n-1,h, not(h,t)) ; S(1,h,t) ; S(n-1,not(h,t),t)

where n denotes the number of disks to be moved, h denotes the home rod, t denotes the target rod, not(h,t) denotes the third rod (neither h nor t), ";" denotes concatenation, and

: S(n, h, t) := solution to a problem consisting of n disks that are to be moved from rod h to rod t.

For n=1 the problem is trivial, namely S(1,h,t) = "move a disk from rod h to rod t" (there is only one disk left).

The  number of moves required by this solution is 2<sup>''n''</sup>&nbsp;&minus;&nbsp;1. If the objective is to '''maximize''' the number of moves (without cycling) then the dynamic programming [[Bellman equation|functional equation]]  is slightly more complicated and  3<sup>''n''</sup>&nbsp;&minus;&nbsp;1 moves are required.<ref>{{Citation |author=Moshe Sniedovich |title= OR/MS Games: 2. The Towers of Hanoi Problem |journal=INFORMS Transactions on Education |volume=3 |issue=1 |year=2002 |pages=34–51 |doi= 10.1287/ited.3.1.45 |postscript=.|doi-access=free }}</ref>

=== Egg dropping puzzle ===
The following is a description of the instance of this famous [[puzzle]] involving N=2 eggs and a building with H=36 floors:<ref>Konhauser J.D.E., Velleman, D., and Wagon, S. (1996). [https://books.google.com/books?id=ElSi5V5uS2MC Which way did the Bicycle Go?] Dolciani Mathematical Expositions – No 18.  [[The Mathematical Association of America]].</ref> 
:Suppose that we wish to know which stories in a 36-story building are safe to drop eggs from, and which will cause the eggs to break on landing (using [[U.S. English]] terminology, in which the first floor is at ground level). We make a few assumptions:

:* An egg that survives a fall can be used again.
:* A broken egg must be discarded.
:* The effect of a fall is the same for all eggs.
:* If an egg breaks when dropped, then it would break if dropped from a higher window.
:* If an egg survives a fall, then it would survive a shorter fall.
:* It is not ruled out that the first-floor windows break eggs, nor is it ruled out that eggs can survive the 36th-floor windows.

: If only one egg is available and we wish to be sure of obtaining the right result, the experiment can be carried out in only one way. Drop the egg from the first-floor window; if it survives, drop it from the second-floor window. Continue upward until it breaks. In the worst case, this method may require 36 droppings. Suppose 2 eggs are available. What is the lowest number of egg-droppings that is guaranteed to work in all cases?

To derive a dynamic programming [[Bellman equation|functional equation]] for this puzzle, let the '''state''' of the dynamic programming model be a pair s = (n,k), where

: ''n'' = number of test eggs available, ''n'' = 0, 1, 2, 3, ..., ''N''&nbsp;&minus;&nbsp;1.
: ''k'' = number of (consecutive) floors yet to be tested, ''k'' = 0, 1, 2, ..., ''H''&nbsp;&minus;&nbsp;1.

For instance, ''s'' = (2,6) indicates that two test eggs are available and 6 (consecutive) floors are yet to be tested. The initial state of the process is ''s'' = (''N'',''H'') where ''N'' denotes the number of test eggs available at the commencement of the experiment. The process terminates either when there are no more test eggs (''n'' = 0) or when ''k'' = 0, whichever occurs first. If termination occurs at state ''s'' = (0,''k'') and ''k''&nbsp;>&nbsp;0, then the test failed.

Now, let

: ''W''(''n'',''k'') = minimum number of trials required to identify the value of the critical floor under the worst-case scenario given that the process is in state ''s'' = (''n'',''k'').

Then it can be shown that<ref name="sniedovich_03">{{Cite journal|doi = 10.1287/ited.4.1.48|title = OR/MS Games: 4. The Joy of Egg-Dropping in Braunschweig and Hong Kong|year = 2003|last1 = Sniedovich| first1 = Moshe|journal = INFORMS Transactions on Education|volume = 4| issue=1 |pages = 48–64|doi-access = free}}</ref>

: ''W''(''n'',''k'') = 1 + min{max(''W''(''n'' &minus; 1, ''x'' &minus; 1), ''W''(''n'',''k'' &minus; ''x'')): ''x'' = 1, 2, ..., ''k'' }

with ''W''(''n'',0) = 0 for all ''n''&nbsp;>&nbsp;0 and ''W''(1,''k'') = ''k'' for all&nbsp;''k''. It is easy to solve this equation iteratively by systematically increasing the values of ''n'' and&nbsp;''k''.

==== Faster DP solution using a different parametrization ====
Notice that the above solution takes <math>O( n k^2 )</math> time with a DP solution. This can be improved to <math>O( n k \log k )</math> time by binary searching on the optimal <math>x</math> in the above recurrence, since <math>W(n-1,x-1)</math> is increasing in <math>x</math> while <math>W(n,k-x)</math> is decreasing in <math>x</math>, thus a local minimum of <math>\max(W(n-1,x-1),W(n,k-x))</math> is a global minimum. Also, by storing the optimal <math>x</math> for each cell in the DP table and referring to its value for the previous cell, the optimal <math>x</math> for each cell can be found in constant time, improving it to <math>O( n k )</math> time. However, there is an even faster solution that involves a different parametrization of the problem:

Let <math>k</math> be the total number of floors such that the eggs break when dropped from the <math>k</math>th floor (The example above is equivalent to taking <math>k=37</math>).

Let <math>m</math> be the minimum floor from which the egg must be dropped to be broken.

Let <math>f(t,n)</math> be the maximum number of values of <math>m</math> that are distinguishable using <math>t</math> tries and <math>n</math> eggs.

Then <math>f(t,0) = f(0,n) = 1</math> for all <math>t,n \geq 0</math>.

Let <math>a</math> be the floor from which the first egg is dropped in the optimal strategy.

If the first egg broke, <math>m</math> is from <math>1</math> to <math>a</math> and distinguishable using at most <math>t-1</math> tries and <math>n-1</math> eggs.

If the first egg did not break, <math>m</math> is from <math>a+1</math> to <math>k</math> and distinguishable using <math>t-1</math> tries and <math>n</math> eggs.

Therefore, <math>f(t,n) = f(t-1,n-1) + f(t-1,n)</math>.

Then the problem is equivalent to finding the minimum <math>x</math> such that <math>f(x,n) \geq k</math>.

To do so, we could compute <math>\{ f(t,i) : 0 \leq i \leq n \}</math> in order of increasing <math>t</math>, which would take <math>O( n x )</math> time.

Thus, if we separately handle the case of <math>n=1</math>, the algorithm would take <math>O( n \sqrt{k} )</math> time.

But the recurrence relation can in fact be solved, giving <math>f(t,n) = \sum_{i=0}^{n}{ \binom{t}{i} }</math>, which can be computed in <math>O(n)</math> time using the identity <math>\binom{t}{i+1} = \binom{t}{i} \frac{t-i}{i+1}</math> for all <math>i \geq 0</math>.

Since <math>f(t,n) \leq f(t+1,n)</math> for all <math>t \geq 0</math>, we can binary search on <math>t</math> to find <math>x</math>, giving an <math>O( n \log k )</math> algorithm.<ref>{{Citation |author=Dean Connable Wills |title=Connections between combinatorics of permutations and algorithms and geometry |url=https://ir.library.oregonstate.edu/xmlui/handle/1957/11929?show=full}}</ref>

=== Matrix chain multiplication ===
{{unreferenced section|date=May 2013}}
{{Main|Matrix chain multiplication}}
<!--Show how the placement of parentheses affects the number of scalar multiplications required when multiplying a bunch of matrices.
Show how to write a dynamic program to calculate the optimal parentheses placement.
This is such a long example that it might be better to make it its own article.-->

Matrix chain multiplication is a well-known example that demonstrates utility of dynamic programming. For example,  engineering applications often have to multiply a chain of matrices. It is not surprising to find matrices of large dimensions, for example 100×100. Therefore, our task is to multiply matrices {{tmath|A_1, A_2, .... A_n}}. Matrix multiplication is not commutative, but is associative; and we can multiply only two matrices at a time. So, we can multiply this chain of matrices in many different ways, for example:

: {{math|((A<sub>1</sub> × A<sub>2</sub>) × A<sub>3</sub>) × ... A<sub>n</sub>}}

: {{math|A<sub>1</sub>×(((A<sub>2</sub>×A<sub>3</sub>)× ... ) × A<sub>n</sub>)}}

: {{math|(A<sub>1</sub> × A<sub>2</sub>) × (A<sub>3</sub> × ... A<sub>n</sub>)}}

and so on. There are numerous ways to multiply this chain of matrices. They will all produce the same final result, however they will take more or less time to compute, based on which particular matrices are multiplied. If matrix A has dimensions m×n and matrix B has dimensions n×q, then matrix C=A×B will have dimensions m×q, and will require m*n*q scalar multiplications (using a simplistic [[matrix multiplication algorithm]] for purposes of illustration).

For example, let us multiply matrices A, B and C. Let us assume that their dimensions are m×n, n×p, and p×s, respectively. Matrix A×B×C will be of size m×s and can be calculated in two ways shown below:

# Ax(B×C)    This order of matrix multiplication will require nps + mns scalar multiplications.  
# (A×B)×C    This order of matrix multiplication will require mnp + mps scalar calculations.

Let us assume that m = 10, n = 100, p = 10 and s = 1000. So, the first way to multiply the chain will require 1,000,000 + 1,000,000 calculations. The second way will require only 10,000+100,000 calculations. Obviously, the second way is faster, and we should multiply the matrices using that arrangement of parenthesis.

Therefore, our conclusion is that the order of parenthesis matters, and that our task is to find the optimal order of parenthesis.

At this point, we have several choices, one of which is to design a dynamic programming algorithm that will split the problem into overlapping problems and calculate the optimal arrangement of parenthesis. The dynamic programming solution is presented below.

Let's call m[i,j] the minimum number of scalar multiplications needed to multiply a chain of matrices from matrix i to matrix j (i.e. A<sub>i</sub> × .... × A<sub>j</sub>, i.e. i<=j). We split the chain at some matrix k, such that i <= k < j, and try to find out which combination produces minimum m[i,j].

The formula is:

        '''if''' i = j, m[i,j]= 0
        '''if''' i < j, m[i,j]= min over all possible values of k {{nowrap|(m[i,k]+m[k+1,j] + <math>p_{i-1}*p_k*p_j</math>)}} 
where ''k'' ranges from ''i'' to ''j''&nbsp;&minus;&nbsp;1.

*{{tmath|p_{{(}}i-1{{)}}}} is the row dimension of matrix i,
*{{tmath|p_k}} is the column dimension of matrix k,
*{{tmath|p_j}} is the column dimension of matrix j.

This formula can be coded as shown below, where input parameter "chain" is the chain of matrices, i.e. {{tmath|A_1, A_2, ... A_n}}:

 '''function''' OptimalMatrixChainParenthesis(chain)
     n = length(chain)
     '''for''' i = 1, n
         m[i,i] = 0    ''// Since it takes no calculations to multiply one matrix''
     '''for''' len = 2, n
         '''for''' i = 1, n - len + 1
             j = i + len -1
             m[i,j] = infinity      ''// So that the first calculation updates''
             '''for''' k = i, j-1
                 {{nowrap|1=q = m[i, k] + m[k+1, j] + <math>p_{i-1}*p_k*p_j</math>}}
                 '''if''' q < m[i, j]     ''// The new order of parentheses is better than what we had''
                     m[i, j] = q    ''// Update''
                     s[i, j] = k    ''// Record which k to split on, i.e. where to place the parenthesis''

So far, we have calculated values for all possible {{math|''m''[''i'', ''j'']}}, the minimum number of calculations to multiply a chain from matrix ''i'' to matrix ''j'', and we have recorded the corresponding "split point"{{math|''s''[''i'', ''j'']}}. For example, if we are multiplying chain {{math|A<sub>1</sub>×A<sub>2</sub>×A<sub>3</sub>×A<sub>4</sub>}}, and it turns out that {{math|1=''m''[1, 3] = 100}} and {{math|1=''s''[1, 3] = 2}}, that means that the optimal placement of parenthesis for matrices 1 to 3 is {{tmath|(A_1\times A_2)\times A_3}} and to multiply those matrices will require 100 scalar calculations.

This algorithm will produce "tables" ''m''[, ] and ''s''[, ] that will have entries for all possible values of i and j. The final solution for the entire chain is m[1, n], with corresponding split at s[1, n]. Unraveling the solution will be recursive, starting from the top and continuing until we reach the base case, i.e. multiplication of single matrices.

Therefore, the next step is to actually split the chain, i.e. to place the parenthesis where they (optimally) belong. For this purpose we could use the following algorithm:

 '''function''' PrintOptimalParenthesis(s, i, j)
     '''if''' i = j
         print "A"i
     '''else'''
         print "(" 
         PrintOptimalParenthesis(s, i, s[i, j]) 
         PrintOptimalParenthesis(s, s[i, j] + 1, j) 
         print ")"

Of course, this algorithm is not useful for actual multiplication. This algorithm is just a user-friendly way to see what the result looks like.

To actually multiply the matrices using the proper splits, we need the following algorithm:
<syntaxhighlight lang="javascript">
   function MatrixChainMultiply(chain from 1 to n)       // returns the final matrix, i.e. A1×A2×... ×An
      OptimalMatrixChainParenthesis(chain from 1 to n)   // this will produce s[ . ] and m[ . ] "tables"
      OptimalMatrixMultiplication(s, chain from 1 to n)  // actually multiply

   function OptimalMatrixMultiplication(s, i, j)   // returns the result of multiplying a chain of matrices from Ai to Aj in optimal way
      if i < j
         // keep on splitting the chain and multiplying the matrices in left and right sides
         LeftSide = OptimalMatrixMultiplication(s, i, s[i, j])
         RightSide = OptimalMatrixMultiplication(s, s[i, j] + 1, j)
         return MatrixMultiply(LeftSide, RightSide) 
      else if i = j
         return Ai   // matrix at position i
      else 
         print "error, i <= j must hold"

    function MatrixMultiply(A, B)    // function that multiplies two matrices
      if columns(A) = rows(B) 
         for i = 1, rows(A)
            for j = 1, columns(B)
               C[i, j] = 0
               for k = 1, columns(A)
                   C[i, j] = C[i, j] + A[i, k]*B[k, j] 
               return C 
      else 
          print "error, incompatible dimensions."
</syntaxhighlight>