Introduction Radboud University Nijmegen

Dynamic programming II

Alexandra Silva
[email protected]
http://www.cs.ru.nl/~alexandra

Institute for Computing and Information Sciences

Radboud University Nijmegen

7th October 2014

Alexandra 7th October 2014 Lesson 4 1 / 19

 Introduction Radboud University Nijmegen
Recap

Message of last week

• Simple, powerful technique.
• Particularly suited for optimisation problems (max, min, etc).
• Two examples: Fibonacci and Shortest paths (Bellman-Ford).
• This week: DP in 5 steps; three more examples (text
justification, black jack, invariant sets).

Alexandra 7th October 2014 Lesson 4 2 / 19

 Introduction Radboud University Nijmegen
Dynamic programming

DP = careful brute force

DP = guessing + memoize + recursion

DP = shortest path in a DAG

Total time = number of sub-problems × time per sub-problem

Cool thing: recursion is for free (constant time)

Alexandra 7th October 2014 Lesson 4 3 / 19

 Introduction Radboud University Nijmegen
Dynamic programming in 5 steps

• Steps are not necessarily sequential.

• How to express a DP, how to invent one, how to explain one.

1 Define sub-problems.
2 Guess (part of solution).
3 Relate sub-problem solutions (recurrence).
4 Build an algorithm: memoization (or bottom up build the DP
table).
5 Solve original problem.

Alexandra 7th October 2014 Lesson 4 4 / 19

 Introduction Radboud University Nijmegen
Last week’s examples
Fibonnacci Shortest Path
1 f1 , . . . , fn δk (s, v ) for v ∈ V , 0 ≤ k ≤ |V |
n.o subp n |V |2

Alexandra 7th October 2014 Lesson 4 5 / 19

 Introduction Radboud University Nijmegen
Last week’s examples
Fibonnacci Shortest Path
1 f1 , . . . , fn δk (s, v ) for v ∈ V , 0 ≤ k ≤ |V |
n.o subp n |V |2

2 nothing edge into v

n.o choices - in-degree(v) +1

Alexandra 7th October 2014 Lesson 4 5 / 19

 Introduction Radboud University Nijmegen
Last week’s examples
Fibonnacci Shortest Path
1 f1 , . . . , fn δk (s, v ) for v ∈ V , 0 ≤ k ≤ |V |
n.o subp n |V |2

2 nothing edge into v

n.o choices - in-degree(v) +1
3 Fk = Fk−1 + Fk−2 δk (s, v ) = min{δk−1 (s, u) + w (u, v )

| (u, v ) ∈ E }
time per subp constant in-degree(v) +1

Alexandra 7th October 2014 Lesson 4 5 / 19

 Introduction Radboud University Nijmegen
Last week’s examples
Fibonnacci Shortest Path
1 f1 , . . . , fn δk (s, v ) for v ∈ V , 0 ≤ k ≤ |V |
n.o subp n |V |2

2 nothing edge into v

n.o choices - in-degree(v) +1
3 Fk = Fk−1 + Fk−2 δk (s, v ) = min{δk−1 (s, u) + w (u, v )

| (u, v ) ∈ E }
time per subp constant in-degree(v) +1

4 for k = 1, . . . , n for k = 0, 1, . . . , |V |
for v ∈ V
topological DAG

Alexandra 7th October 2014 Lesson 4 5 / 19

 Introduction Radboud University Nijmegen
Last week’s examples
Fibonnacci Shortest Path
1 f1 , . . . , fn δk (s, v ) for v ∈ V , 0 ≤ k ≤ |V |
n.o subp n |V |2

2 nothing edge into v

n.o choices - in-degree(v) +1
3 Fk = Fk−1 + Fk−2 δk (s, v ) = min{δk−1 (s, u) + w (u, v )

| (u, v ) ∈ E }
time per subp constant in-degree(v) +1

4 for k = 1, . . . , n for k = 0, 1, . . . , |V |
for v ∈ V
topological DAG

5 Fn δ|V |−1 (s, v )(no negative weight cycles!)

extra time constant O(|V |)
Alexandra 7th October 2014 Lesson 4 5 / 19
 Introduction Radboud University Nijmegen
Text justification

Given some text (a string), split it into good lines.

Rules:
• We can only split between word boundaries.
• Write some text (with spaces in it), new line, more text, etc.
• Justified text (not ugly ;-)).

some words in one line

text and|{z}more text
ugly

Alexandra 7th October 2014 Lesson 4 6 / 19

 Introduction Radboud University Nijmegen
Text justification

Given some text (a string), split it into good lines.

Rules:
• We can only split between word boundaries.
• Write some text (with spaces in it), new line, more text, etc.
• Justified text (not ugly ;-)).

some words in one line

text and|{z}more text
ugly

• MS word (older versions): greedy strategy, pack as many

words on the first line, etc.
• LaTeX: DP, what we will see here.

Alexandra 7th October 2014 Lesson 4 6 / 19

 Introduction Radboud University Nijmegen
Text justification

• Text: list of words.

• We define badness(i,j) – ugliness measure: how bad it is to
use words i through j as a line.

Alexandra 7th October 2014 Lesson 4 7 / 19

 Introduction Radboud University Nijmegen
Text justification

• Text: list of words.

• We define badness(i,j) – ugliness measure: how bad it is to
use words i through j as a line.
• Two options: words fit on a line or they do not fit.

Alexandra 7th October 2014 Lesson 4 7 / 19

 Introduction Radboud University Nijmegen
Text justification

• Text: list of words.

• We define badness(i,j) – ugliness measure: how bad it is to
use words i through j as a line.
• Two options: words fit on a line or they do not fit.
• Sum of lengths plus spaces (minimised) larger than width of
page = bad (∞).
• Otherwise = (page width minus total width)3 . (why cubed?)

Alexandra 7th October 2014 Lesson 4 7 / 19

 Introduction Radboud University Nijmegen
Text justification

• Text: list of words.

• We define badness(i,j) – ugliness measure: how bad it is to
use words i through j as a line.
• Two options: words fit on a line or they do not fit.
• Sum of lengths plus spaces (minimised) larger than width of
page = bad (∞).
• Otherwise = (page width minus total width)3 . (why cubed?)
• total and page width very close = difference small = cubed
value very small
• total and page width not close = difference large = cubed
value very large = discourages huge spaces

Alexandra 7th October 2014 Lesson 4 7 / 19

 Introduction Radboud University Nijmegen
Text justification

• Greedy algorithm: make the first line very good but might
misbalance all the rest.
• Overall goal: minimise sum of badness of all lines.
• Sum of cubes: looks hard. . .

Alexandra 7th October 2014 Lesson 4 8 / 19

 Introduction Radboud University Nijmegen
Text justification

• Greedy algorithm: make the first line very good but might
misbalance all the rest.
• Overall goal: minimise sum of badness of all lines.
• Sum of cubes: looks hard. . . DP helps! You don’t need to
think ;-)

Alexandra 7th October 2014 Lesson 4 8 / 19

 Introduction Radboud University Nijmegen
Text justification

• Define sub-problems (hard part).

• What is the brute-force strategy?
• Try all possibilities?

Alexandra 7th October 2014 Lesson 4 9 / 19

 Introduction Radboud University Nijmegen
Text justification

• Define sub-problems (hard part).

• What is the brute-force strategy?
• Try all possibilities?

• Try all partitions: it could fit on one line, maybe two (but
then where do I split?)
• For every word: start a new line or not?
• There are n words: 2n different splittings.

Alexandra 7th October 2014 Lesson 4 9 / 19

 Introduction Radboud University Nijmegen
Text justification

• Define sub-problems (hard part).

• What is the brute-force strategy?
• Try all possibilities?

• Try all partitions: it could fit on one line, maybe two (but
then where do I split?)
• For every word: start a new line or not?
• There are n words: 2n different splittings.
• But where do the lines begin?

Alexandra 7th October 2014 Lesson 4 9 / 19

 Introduction Radboud University Nijmegen
Text justification

• Define sub-problems (hard part).

• What is the brute-force strategy?
• Try all possibilities?

• Try all partitions: it could fit on one line, maybe two (but
then where do I split?)
• For every word: start a new line or not?
• There are n words: 2n different splittings.
• But where do the lines begin?
• What can I (first) guess when given the huge string of words?

Alexandra 7th October 2014 Lesson 4 9 / 19

 Introduction Radboud University Nijmegen
Text justification

• Define sub-problems (hard part).

• What is the brute-force strategy?
• Try all possibilities?

• Try all partitions: it could fit on one line, maybe two (but
then where do I split?)
• For every word: start a new line or not?
• There are n words: 2n different splittings.
• But where do the lines begin?
• What can I (first) guess when given the huge string of words?
• Guess how long first line is (=where the second line begins).

Alexandra 7th October 2014 Lesson 4 9 / 19

 Introduction Radboud University Nijmegen
Text justification

• Try all words to begin second line.

• After I know where the second line begins. . .

Alexandra 7th October 2014 Lesson 4 10 / 19

 Introduction Radboud University Nijmegen
Text justification

• Try all words to begin second line.

• After I know where the second line begins. . . I need to know
where the third begins!

Alexandra 7th October 2014 Lesson 4 10 / 19

 Introduction Radboud University Nijmegen
Text justification

• Try all words to begin second line.

• After I know where the second line begins. . . I need to know
where the third begins!
• And then the fourth, and then . . .

Alexandra 7th October 2014 Lesson 4 10 / 19

 Introduction Radboud University Nijmegen
Text justification

• Try all words to begin second line.

• After I know where the second line begins. . . I need to know
where the third begins!
• And then the fourth, and then . . .
• Sub-problems: after first guess problem of original type!

Alexandra 7th October 2014 Lesson 4 10 / 19

 Introduction Radboud University Nijmegen
Text justification

• Try all words to begin second line.

• After I know where the second line begins. . . I need to know
where the third begins!
• And then the fourth, and then . . .
• Sub-problems: after first guess problem of original type!
• Originally: all words. After first guess: remaining words.
• Sub-problems: suffixes.
• How many?

Alexandra 7th October 2014 Lesson 4 10 / 19

 Introduction Radboud University Nijmegen
Text justification

• Try all words to begin second line.

• After I know where the second line begins. . . I need to know
where the third begins!
• And then the fourth, and then . . .
• Sub-problems: after first guess problem of original type!
• Originally: all words. After first guess: remaining words.
• Sub-problems: suffixes.
• How many?
• Key observation: If we think for every word: in or out then 2n .
But in fact we do that for remaining words.
• First line: 100 words, remaining: n − 100; First line: 2000
words, remaining: n − 2000 → n possibilities.

Alexandra 7th October 2014 Lesson 4 10 / 19

 Introduction Radboud University Nijmegen
Text justification

• Try all words to begin second line.

• After I know where the second line begins. . .

Alexandra 7th October 2014 Lesson 4 11 / 19

 Introduction Radboud University Nijmegen
Text justification

• Try all words to begin second line.

• After I know where the second line begins. . . I need to know
where the third begins!

Alexandra 7th October 2014 Lesson 4 11 / 19

 Introduction Radboud University Nijmegen
Text justification

• Try all words to begin second line.

• After I know where the second line begins. . . I need to know
where the third begins!
• And then the fourth, and then . . .

Alexandra 7th October 2014 Lesson 4 11 / 19

 Introduction Radboud University Nijmegen
Text justification

• Try all words to begin second line.

• After I know where the second line begins. . . I need to know
where the third begins!
• And then the fourth, and then . . .
• Sub-problems: after first guess problem of original type!

Alexandra 7th October 2014 Lesson 4 11 / 19

 Introduction Radboud University Nijmegen
Text justification

• Try all words to begin second line.

• After I know where the second line begins. . . I need to know
where the third begins!
• And then the fourth, and then . . .
• Sub-problems: after first guess problem of original type!
• Originally: all words. After first guess: remaining words.
• Sub-problems: suffixes.
• How many?

Alexandra 7th October 2014 Lesson 4 11 / 19

 Introduction Radboud University Nijmegen
Text justification

• Try all words to begin second line.

• After I know where the second line begins. . . I need to know
where the third begins!
• And then the fourth, and then . . .
• Sub-problems: after first guess problem of original type!
• Originally: all words. After first guess: remaining words.
• Sub-problems: suffixes.
• How many?
• Key observation: If we think for every word: in or out then 2n .
But in fact we do that for remaining words.
• First line: 100 words, remaining: n − 100; First line: 2000
words, remaining: n − 2000 → n possibilities.

Alexandra 7th October 2014 Lesson 4 11 / 19

 Introduction Radboud University Nijmegen
Text justification

1 subproblems : suffixes ; number: n.

Given some words (n − i) how do we justify the text? )

Alexandra 7th October 2014 Lesson 4 12 / 19

 Introduction Radboud University Nijmegen
Text justification

1 subproblems : suffixes ; number: n.

Given some words (n − i) how do we justify the text? )
2 guess: where to start the second line; number: ≤ n − i ≈ O(n)

Alexandra 7th October 2014 Lesson 4 12 / 19

 Introduction Radboud University Nijmegen
Text justification

1 subproblems : suffixes ; number: n.

Given some words (n − i) how do we justify the text? )
2 guess: where to start the second line; number: ≤ n − i ≈ O(n)
3 recurrrence: TJ(i) = ?
we have done i-1 lines
i some words and so
j some other words
And a lot of more stuff

Alexandra 7th October 2014 Lesson 4 12 / 19

 Introduction Radboud University Nijmegen
Text justification

1 subproblems : suffixes ; number: n.

Given some words (n − i) how do we justify the text? )
2 guess: where to start the second line; number: ≤ n − i ≈ O(n)
3 recurrrence: TJ(i) = ?
we have done i-1 lines
i some words and so
j some other words
And a lot of more stuff
TJ(i) =

Alexandra 7th October 2014 Lesson 4 12 / 19

 Introduction Radboud University Nijmegen
Text justification

1 subproblems : suffixes ; number: n.

Given some words (n − i) how do we justify the text? )
2 guess: where to start the second line; number: ≤ n − i ≈ O(n)
3 recurrrence: TJ(i) = ?
we have done i-1 lines
i some words and so
j some other words
And a lot of more stuff
TJ(i) = TJ(j) +

Alexandra 7th October 2014 Lesson 4 12 / 19

 Introduction Radboud University Nijmegen
Text justification

1 subproblems : suffixes ; number: n.

Given some words (n − i) how do we justify the text? )
2 guess: where to start the second line; number: ≤ n − i ≈ O(n)
3 recurrrence: TJ(i) = ?
we have done i-1 lines
i some words and so
j some other words
And a lot of more stuff
TJ(i) = TJ(j) + badness(i,j)

Alexandra 7th October 2014 Lesson 4 12 / 19

 Introduction Radboud University Nijmegen
Text justification

1 subproblems : suffixes ; number: n.

Given some words (n − i) how do we justify the text? )
2 guess: where to start the second line; number: ≤ n − i ≈ O(n)
3 recurrrence: TJ(i) = ?
we have done i-1 lines
i some words and so
j some other words
And a lot of more stuff
TJ(i) = TJ(j) + badness(i,j) for j = i + 1 . . . n + 1

Alexandra 7th October 2014 Lesson 4 12 / 19

 Introduction Radboud University Nijmegen
Text justification

1 subproblems : suffixes ; number: n.

Given some words (n − i) how do we justify the text? )
2 guess: where to start the second line; number: ≤ n − i ≈ O(n)
3 recurrrence: TJ(i) = ?
we have done i-1 lines
i some words and so
j some other words
And a lot of more stuff
TJ(i) = min{TJ(j) + badness(i,j) for j = i + 1 . . . n + 1 }
Time per subproblem =

Alexandra 7th October 2014 Lesson 4 12 / 19

 Introduction Radboud University Nijmegen
Text justification

1 subproblems : suffixes ; number: n.

Given some words (n − i) how do we justify the text? )
2 guess: where to start the second line; number: ≤ n − i ≈ O(n)
3 recurrrence: TJ(i) = ?
we have done i-1 lines
i some words and so
j some other words
And a lot of more stuff
TJ(i) = min{TJ(j) + badness(i,j) for j = i + 1 . . . n + 1 }
Time per subproblem = O(n).

Alexandra 7th October 2014 Lesson 4 12 / 19

 Introduction Radboud University Nijmegen
Text justification

4 Topological order:

Alexandra 7th October 2014 Lesson 4 13 / 19

 Introduction Radboud University Nijmegen
Text justification

4 Topological order: n, n − 1, . . . , 0 (Base case:

Alexandra 7th October 2014 Lesson 4 13 / 19

 Introduction Radboud University Nijmegen
Text justification

4 Topological order: n, n − 1, . . . , 0 (Base case: TJ(n) = 0)

Total time =

Alexandra 7th October 2014 Lesson 4 13 / 19

 Introduction Radboud University Nijmegen
Text justification

4 Topological order: n, n − 1, . . . , 0 (Base case: TJ(n) = 0)

Total time = O(n2 )
5 Original problem:

Alexandra 7th October 2014 Lesson 4 13 / 19

 Introduction Radboud University Nijmegen
Text justification

4 Topological order: n, n − 1, . . . , 0 (Base case: TJ(n) = 0)

Total time = O(n2 )
5 Original problem: TJ(0).

Alexandra 7th October 2014 Lesson 4 13 / 19

 Introduction Radboud University Nijmegen
Text justification

4 Topological order: n, n − 1, . . . , 0 (Base case: TJ(n) = 0)

Total time = O(n2 )
5 Original problem: TJ(0).

• This does not return how to pack the words! Just gives you
the minimum badness.
• Similar to shortest path.

Alexandra 7th October 2014 Lesson 4 13 / 19

 Introduction Radboud University Nijmegen
Text justification

4 Topological order: n, n − 1, . . . , 0 (Base case: TJ(n) = 0)

Total time = O(n2 )
5 Original problem: TJ(0).

• This does not return how to pack the words! Just gives you
the minimum badness.
• Similar to shortest path.
• Parent pointer: remember which guess was best.
parent[i] = argmin(. . .) = j value
To reconstruct where the lines break:
0 → parent[0] → parent[parent[0]] → . . .

Alexandra 7th October 2014 Lesson 4 13 / 19

 Introduction Radboud University Nijmegen
Text justification

4 Topological order: n, n − 1, . . . , 0 (Base case: TJ(n) = 0)

Total time = O(n2 )
5 Original problem: TJ(0).

• This does not return how to pack the words! Just gives you
the minimum badness.
• Similar to shortest path.
• Parent pointer: remember which guess was best.
parent[i] = argmin(. . .) = j value
To reconstruct where the lines break:
0 → parent[0] → parent[parent[0]] → . . .
• Total automatic technique (like memoisation).

Alexandra 7th October 2014 Lesson 4 13 / 19

 Introduction Radboud University Nijmegen
Invariant Sets (TC628 Round 1 - Division II, Level Three)

Janusz is a young physicist. Recently he has been studying a dynamic

system. There are n elements in the system. We will label them 0
through n − 1. The set of all elements will be denoted E . Janusz models
the changes in his system using a function that operates on elements of
E . You are given this function as a list of integers f with n elements,
each between 0 and n − 1, inclusive. The list f describes Janusz’s
function as follows: For each valid i, if the function is given the input i,
its output will be f [i]. For Janusz, invariant subsets of E have a special
significance. A subset S of the set E is called invariant (with respect to
f ) if it has the following property: for any x ∈ S, f [x] is also in S. For
example, the entire set E is invariant (for any f ). The empty set is
always invariant as well.
Given is the list f , compute and return the number of invariant subsets of
E (including E itself and the empty set).

Alexandra 7th October 2014 Lesson 4 14 / 19

 Introduction Radboud University Nijmegen
Invariant Sets
{1, 0, 0, 0}
Returns: 5
The invariant sets are: {}, {0, 1}, {0, 1, 2}, {0, 1, 3}, {0, 1, 2, 3}.

{1, 2, 0}
Returns: 2
There are only 2 invariants sets: {} and {0, 1, 2}.

{0, 0, 1, 2}
Returns: 5
The invariant sets are: {}, {0}, {0, 1}, {0, 1, 2}, {0, 1, 2, 3}.

{0, 1, 2, 3, 4, 5}
Returns: 64; Every set is invariant when f(x)=x for all x.

{12, 10, 0, 4, 0, 6, 3, 8, 9, 14, 1, 5, 6, 12, 5}

Returns: 90
Alexandra 7th October 2014 Lesson 4 15 / 19
 Introduction Radboud University Nijmegen
Blackjack

• Dealer+player.
• One card of the dealer is hidden.
• Player decides where to stander hit.

Alexandra 7th October 2014 Lesson 4 16 / 19

 Introduction Radboud University Nijmegen
Blackjack

• Perfect-information blackjack (inside man!).

• Knowledge of the whole deck ahead of time.
• At each stage, decide whether to hit or stand.
• Similar strategy to text justification.

Alexandra 7th October 2014 Lesson 4 17 / 19

 Introduction Radboud University Nijmegen
Blackjack
Parameters:
• Deck: c0 , c1 , . . . cn−1
• 1 player vs dealer (solution for two players?); 1 euro bet/hand
1 subproblems:
2 Guess:

Alexandra 7th October 2014 Lesson 4 18 / 19

 Introduction Radboud University Nijmegen
Blackjack
Parameters:
• Deck: c0 , c1 , . . . cn−1
• 1 player vs dealer (solution for two players?); 1 euro bet/hand
1 subproblems:
2 Guess: how many hits?

Alexandra 7th October 2014 Lesson 4 18 / 19

 Introduction Radboud University Nijmegen
Blackjack
Parameters:
• Deck: c0 , c1 , . . . cn−1
• 1 player vs dealer (solution for two players?); 1 euro bet/hand
1 subproblems: suffixes: cards left with
2 Guess: how many hits?

Alexandra 7th October 2014 Lesson 4 18 / 19

 Introduction Radboud University Nijmegen
Blackjack
Parameters:
• Deck: c0 , c1 , . . . cn−1
• 1 player vs dealer (solution for two players?); 1 euro bet/hand
1 subproblems: suffixes: cards left with
2 Guess: how many hits?
Number of subproblems:

Alexandra 7th October 2014 Lesson 4 18 / 19

 Introduction Radboud University Nijmegen
Blackjack
Parameters:
• Deck: c0 , c1 , . . . cn−1
• 1 player vs dealer (solution for two players?); 1 euro bet/hand
1 subproblems: suffixes: cards left with
2 Guess: how many hits?
Number of subproblems: n (suffixes! not subsets!).
Number of choices:

Alexandra 7th October 2014 Lesson 4 18 / 19

 Introduction Radboud University Nijmegen
Blackjack
Parameters:
• Deck: c0 , c1 , . . . cn−1
• 1 player vs dealer (solution for two players?); 1 euro bet/hand
1 subproblems: suffixes: cards left with
2 Guess: how many hits?
Number of subproblems: n (suffixes! not subsets!).
Number of choices: at most n

Alexandra 7th October 2014 Lesson 4 18 / 19

 Introduction Radboud University Nijmegen
Blackjack
Parameters:
• Deck: c0 , c1 , . . . cn−1
• 1 player vs dealer (solution for two players?); 1 euro bet/hand
1 subproblems: suffixes: cards left with
2 Guess: how many hits?
Number of subproblems: n (suffixes! not subsets!).
Number of choices: at most n
3 Recurrence:
BJ(i) =

Alexandra 7th October 2014 Lesson 4 18 / 19

 Introduction Radboud University Nijmegen
Blackjack
Parameters:
• Deck: c0 , c1 , . . . cn−1
• 1 player vs dealer (solution for two players?); 1 euro bet/hand
1 subproblems: suffixes: cards left with
2 Guess: how many hits?
Number of subproblems: n (suffixes! not subsets!).
Number of choices: at most n
3 Recurrence:
BJ(i) = max{

Alexandra 7th October 2014 Lesson 4 18 / 19

 Introduction Radboud University Nijmegen
Blackjack
Parameters:
• Deck: c0 , c1 , . . . cn−1
• 1 player vs dealer (solution for two players?); 1 euro bet/hand
1 subproblems: suffixes: cards left with
2 Guess: how many hits?
Number of subproblems: n (suffixes! not subsets!).
Number of choices: at most n
3 Recurrence:
BJ(i) = max{outcome ∈ {−1, 0, 1} +

Alexandra 7th October 2014 Lesson 4 18 / 19

 Introduction Radboud University Nijmegen
Blackjack
Parameters:
• Deck: c0 , c1 , . . . cn−1
• 1 player vs dealer (solution for two players?); 1 euro bet/hand
1 subproblems: suffixes: cards left with
2 Guess: how many hits?
Number of subproblems: n (suffixes! not subsets!).
Number of choices: at most n
3 Recurrence:
BJ(i) = max{outcome ∈ {−1, 0, 1} + BJ(j)

Alexandra 7th October 2014 Lesson 4 18 / 19

 Introduction Radboud University Nijmegen
Blackjack
Parameters:
• Deck: c0 , c1 , . . . cn−1
• 1 player vs dealer (solution for two players?); 1 euro bet/hand
1 subproblems: suffixes: cards left with
2 Guess: how many hits?
Number of subproblems: n (suffixes! not subsets!).
Number of choices: at most n
3 Recurrence:
BJ(i) = max{outcome ∈ {−1, 0, 1} + BJ(j) | hits = 0, · · · , n

j = i+4+hits+hitsdealer }

Alexandra 7th October 2014 Lesson 4 18 / 19

 Introduction Radboud University Nijmegen
Blackjack
Parameters:
• Deck: c0 , c1 , . . . cn−1
• 1 player vs dealer (solution for two players?); 1 euro bet/hand
1 subproblems: suffixes: cards left with
2 Guess: how many hits?
Number of subproblems: n (suffixes! not subsets!).
Number of choices: at most n
3 Recurrence:
BJ(i) = max{outcome ∈ {−1, 0, 1} + BJ(j) | hits = 0, · · · , n

j = i+4+hits+hitsdealer }
Time per sub-problem: O(n2 )
4 Topological order:

Alexandra 7th October 2014 Lesson 4 18 / 19

 Introduction Radboud University Nijmegen
Blackjack
Parameters:
• Deck: c0 , c1 , . . . cn−1
• 1 player vs dealer (solution for two players?); 1 euro bet/hand
1 subproblems: suffixes: cards left with
2 Guess: how many hits?
Number of subproblems: n (suffixes! not subsets!).
Number of choices: at most n
3 Recurrence:
BJ(i) = max{outcome ∈ {−1, 0, 1} + BJ(j) | hits = 0, · · · , n

j = i+4+hits+hitsdealer }
Time per sub-problem: O(n2 )
4 Topological order: for i in reversed(range(n))

Alexandra 7th October 2014 Lesson 4 18 / 19

 Introduction Radboud University Nijmegen
Blackjack
Parameters:
• Deck: c0 , c1 , . . . cn−1
• 1 player vs dealer (solution for two players?); 1 euro bet/hand
1 subproblems: suffixes: cards left with
2 Guess: how many hits?
Number of subproblems: n (suffixes! not subsets!).
Number of choices: at most n
3 Recurrence:
BJ(i) = max{outcome ∈ {−1, 0, 1} + BJ(j) | hits = 0, · · · , n

j = i+4+hits+hitsdealer }
Time per sub-problem: O(n2 )
4 Topological order: for i in reversed(range(n))
Total time: O(n3 )

Alexandra 7th October 2014 Lesson 4 18 / 19

 Introduction Radboud University Nijmegen
Blackjack
Parameters:
• Deck: c0 , c1 , . . . cn−1
• 1 player vs dealer (solution for two players?); 1 euro bet/hand
1 subproblems: suffixes: cards left with
2 Guess: how many hits?
Number of subproblems: n (suffixes! not subsets!).
Number of choices: at most n
3 Recurrence:
BJ(i) = max{outcome ∈ {−1, 0, 1} + BJ(j) | hits = 0, · · · , n

j = i+4+hits+hitsdealer }
Time per sub-problem: O(n2 )
4 Topological order: for i in reversed(range(n))
Total time: O(n3 )
5 Solution: BJ(0)
Alexandra 7th October 2014 Lesson 4 18 / 19
 Introduction Radboud University Nijmegen
Blackjack
Recurrence (detailed):

Total time = O(n2 )

Alexandra 7th October 2014 Lesson 4 19 / 19