Dynamic Programming: CSE 431/531: Algorithm Analysis and Design (Spring 2022)

CSE 431/531: Algorithm Analysis and Design (Spring 2022)
Dynamic Programming
Lecturer: Shi Li
Department of Computer Science and Engineering
University at Buffalo
Paradigms for Designing Algorithms
Greedy algorithm
Make a greedy choice
Prove that the greedy choice is safe
Reduce the problem to a sub-problem and solve it iteratively
Usually for optimization problems
Divide-and-conquer
Break a problem into many independent sub-problems
Solve each sub-problem separately
Combine solutions for sub-problems to form a solution for the
original one
Usually used to design more efficient algorithms
2/79
Paradigms for Designing Algorithms
Dynamic Programming
Break up a problem into many overlapping sub-problems
Build solutions for larger and larger sub-problems
Use a table to store solutions for sub-problems for reuse
3/79
Recall: Computing the n-th Fibonacci Number
F0 = 0, F1 = 1
Fn = Fn−1 + Fn−2 , ∀n ≥ 2
Fibonacci sequence: 0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, · · ·
Fib(n)
1: F [0] ← 0
2: F [1] ← 1
3: for i ← 2 to n do
4: F [i] ← F [i − 1] + F [i − 2]
5: return F [n]
Store each F [i] for future use.
4/79
Outline
1 Weighted Interval Scheduling

2 Subset Sum Problem
3 Knapsack Problem
4 Longest Common Subsequence
Longest Common Subsequence in Linear Space
5 Shortest Paths in Directed Acyclic Graphs
6 Matrix Chain Multiplication
7 Optimum Binary Search Tree
8 Summary
5/79
Recall: Interval Schduling
Input: n jobs, job i with start time si and finish time fi
each job has a weight (or value) vi > 0
i and j are compatible if [si , fi ) and [sj , fj ) are disjoint
Output: a maximum-size subset of mutually compatible jobs
0 1 2 3 4 5 6 7 8 9
100 50 30
25 50
90 80
80 70
Optimum value = 220
6/79
Hard to Design a Greedy Algorithm
Q: Which job is safe to schedule?
Job with the earliest finish time? No, we are ignoring weights
Job with the largest weight? No, we are ignoring times
weight
Job with the largest ?
length
No, when weights are equal, this is the shortest job
0 1 2 3 4 5 6 7 8 9
7/79
Designing a Dynamic Programming Algorithm
0 1 2 3 4 5 6 7 8 9
i opt[i]
2 100 5 50 9 30
0 0
4 25 8 50
1 80
3 90 7 80
2 100
1 80 6 70
3 100
4 105
5 150
Sort jobs according to non-decreasing order 6 170
of finish times 7 185
opt[i]: optimal value for instance only 8 220
containing jobs {1, 2, · · · , i} 9 220
8/79
Focus on instance
0 1 2 3 4 5 6 7 8 9
{1, 2, 3, · · · , i},
2 100 5 50 9 30
4 25 8 50
opt[i]: optimal value for the
3 90 7 80
instance
1 80 6 70 assume we have computed
opt[0], opt[1], · · · , opt[i − 1]
Q: The value of optimal solution that does not contain i?
A: opt[i − 1]
Q: The value of optimal solution that contains job i?
A: vi + opt[pi ], pi = the largest j such that fj ≤ si

9/79
Q: The value of optimal solution that does not contain i?
A: opt[i − 1]
Q: The value of optimal solution that contains job i?
A: vi + opt[pi ], pi = the largest j such that fj ≤ si
Recursion for opt[i]:

opt[i] = max {opt[i − 1], vi + opt[pi ]}
10/79

0 1 2 3 4 5 6 7 8 9
2 100 5 50 9 30
4 25 8 50
3 90 7 80
1 80 6 70
opt[0] = 0
opt[1] = max{opt[0], 80 + opt[0]} = 80
opt[2] = max{opt[1], 100 + opt[0]} = 100
opt[3] = max{opt[2], 90 + opt[0]} = 100
opt[4] = max{opt[3], 25 + opt[1]} = 105
opt[5] = max{opt[4], 50 + opt[3]} = 150 11/79

0 1 2 3 4 5 6 7 8 9
2 100 5 50 9 30
4 25 8 50
3 90 7 80
1 80 6 70
opt[0] = 0, opt[1] = 80, opt[2] = 100

opt[3] = 100, opt[4] = 105, opt[5] = 150
opt[6] = max{opt[5], 70 + opt[3]} = 170
opt[7] = max{opt[6], 80 + opt[4]} = 185
opt[8] = max{opt[7], 50 + opt[6]} = 220
opt[9] = max{opt[8], 30 + opt[7]} = 220 12/79
Dynamic Programming
1: sort jobs by non-decreasing order of finishing times

2: compute p1 , p2 , · · · , pn
3: opt[0] ← 0
5: opt[i] ← max{opt[i − 1], vi + opt[pi ]}
Running time sorting: O(n lg n)

Running time for computing p: O(n lg n) via binary search
Running time for computing opt[n]: O(n)
13/79
How Can We Recover the Optimum Schedule?
1: sort jobs by non-decreasing order of

finishing times
1: i ← n, S ← ∅
2: compute p1 , p2 , · · · , pn
2: while i 6= 0 do
3: opt[0] ← 0
3: if b[i] = N then
4: i←i−1
5: if opt[i − 1] ≥ vi + opt[pi ] then
5: else
6: opt[i] ← opt[i − 1]
6: S ← S ∪ {i}
7: b[i] ← N
7: i ← pi
8: else
9: opt[i] ← vi + opt[pi ] 8: return S
10: b[i] ← Y
14/79
Recovering Optimum Schedule: Example
i opt[i] b[i]
0 0 ⊥
1 80 Y
0 1 2 3 4 5 6 7 8 9
2 100 Y
3 100 N 2 100 5 50 9 30
4 105 Y 4 25 8 50
5 150 Y i 3 90 7 80
6 170 Y
1 80 6 70
7 185 Y
8 220 Y
9 220 N
15/79
Dynamic Programming

16/79
Outline

3 Knapsack Problem
8 Summary
17/79
Subset Sum Problem
Input: an integer bound W > 0
a set of n items, each with an integer weight wi > 0
Output: a subset S of items that
X X
maximizes wi s.t. wi ≤ W.
i∈S i∈S
Motivation: you have budget W , and want to buy a subset of

items, so as to spend as much money as possible.
Example:
W = 35, n = 5, w = (14, 9, 17, 10, 13)
Optimum: S = {1, 2, 4} and 14 + 9 + 10 = 33
18/79
Greedy Algorithms for Subset Sum
Candidate Algorithm:
Sort according to non-increasing order of weights
Select items in the order as long as the total weight remains below
W
Q: Does candidate algorithm always produce optimal solutions?
A: No. W = 100, n = 3, w = (51, 50, 50).
Q: What if we change “non-increasing” to “non-decreasing”?
A: No. W = 100, n = 3, w = (1, 50, 50)
19/79
Design a Dynamic Programming Algorithm
Consider the instance: i, W 0 , (w1 , w2 , · · · , wi );

opt[i, W 0 ]: the optimum value of the instance
Q: The value of the optimum solution that does not contain i?
A: opt[i − 1, W 0 ]
Q: The value of the optimum solution that contains i?
A: opt[i − 1, W 0 − wi ] + wi
20/79
Dynamic Programming
Consider the instance: i, W 0 , (w1 , w2 , · · · , wi );

opt[i, W 0 ]: the optimum value of the instance



0 i=0
opt[i − 1, W 0 ]

i > 0, wi > W 0
opt[i, W 0 ] = (
0
)
opt[i − 1, W ]
i > 0, wi ≤ W 0

 max opt[i − 1, W 0 − w ] + w



i i
21/79
Dynamic Programming
1: for W 0 ← 0 to W do
2: opt[0, W 0 ] ← 0
5: opt[i, W 0 ] ← opt[i − 1, W 0 ]
6: if wi ≤ W 0 and opt[i − 1, W 0 − wi ] + wi ≥ opt[i, W 0 ] then
7: opt[i, W 0 ] ← opt[i − 1, W 0 − wi ] + wi
8: return opt[n, W ]
22/79
Recover the Optimum Set
2: opt[0, W 0 ] ← 0
5: opt[i, W 0 ] ← opt[i − 1, W 0 ]
6: b[i, W 0 ] ← N
7: if wi ≤ W 0 and opt[i − 1, W 0 − wi ] + wi ≥ opt[i, W 0 ]
then
8: opt[i, W 0 ] ← opt[i − 1, W 0 − wi ] + wi
9: b[i, W 0 ] ← Y
23/79
Recover the Optimum Set
1: i ← n, W 0 ← W, S ← ∅
2: while i > 0 do
3: if b[i, W 0 ] = Y then
4: W 0 ← W 0 − wi
5: S ← S ∪ {i}
6: i←i−1
7: return S
24/79
Running Time of Algorithm
2: opt[0, W 0 ] ← 0
5: opt[i, W 0 ] ← opt[i − 1, W 0 ]
6: if wi ≤ W 0 and opt[i − 1, W 0 − wi ] + wi ≥ opt[i, W 0 ] then
7: opt[i, W 0 ] ← opt[i − 1, W 0 − wi ] + wi
Running time is O(nW )

Running time is pseudo-polynomial because it depends on value of
the input integers.
25/79
Avoiding Unncessary Computation and Memory
Using Memoized Algorithm and Hash Map
compute-opt(i, W 0 )
1: if opt[i, W 0 ] 6= ⊥ then return opt[i, W 0 ]
2: if i = 0 then r ← 0
3: else
4: r ← compute-opt(i − 1, W 0 )
5: if wi ≤ W 0 then
6: r0 ← compute-opt(i − 1, W 0 − wi ) + wi
7: if r0 > r then r ← r0
8: opt[i, W 0 ] ← r
9: return r
Use hash map for opt

26/79
Outline

3 Knapsack Problem
8 Summary
27/79
Knapsack Problem
Input: an integer bound W > 0
a value vi > 0 for each item i
X X
maximizes vi s.t. wi ≤ W.
i∈S i∈S
Motivation: you have budget W , and want to buy a subset of

items of maximum total value
28/79
DP for Knapsack Problem
opt[i, W 0 ]: the optimum value when budget is W 0 and items are

{1, 2, 3, · · · , i}.
If i = 0, opt[i, W 0 ] = 0 for every W 0 = 0, 1, 2, · · · , W .



 0 i=0
opt[i − 1, W 0 ]

i > 0, wi > W 0
opt[i, W 0 ] = (
0
)
opt[i − 1, W ]
i > 0, wi ≤ W 0

 max opt[i − 1, W 0 − w ] + v



i i
29/79
Exercise: Items with 3 Parameters
Input: integer bounds W > 0, Z > 0,

a size zi > 0 for each item i
a value vi > 0 for each item i
X
maximizes vi s.t.
i∈S
X X
wi ≤ W and zi ≤ Z
i∈S i∈S
30/79
Outline

3 Knapsack Problem
8 Summary
31/79
Subsequence
A = bacdca
C = adca
C is a subsequence of A
Def. Given two sequences A[1 .. n] and C[1 .. t] of letters, C is

called a subsequence of A if there exists integers
1 ≤ i1 < i2 < i3 < . . . < it ≤ n such that A[ij ] = C[j] for every
j = 1, 2, 3, · · · , t.
Exercise: how to check if sequence C is a subsequence of A?
32/79
Longest Common Subsequence
Input: A[1 .. n] and B[1 .. m]
Output: the longest common subsequence of A and B
Example:
A = ‘bacdca0
B = ‘adbcda0
LCS(A, B) = ‘adca0
Applications: edit distance (diff), similarity of DNAs
33/79
Matching View of LCS
b a c d c a
a d b c d a
Goal of LCS: find a maximum-size non-crossing matching between

letters in A and letters in B.
34/79
Reduce to Subproblems
A = ‘bacdca0
B = ‘adbcda0
either the last letter of A is not matched:
need to compute LCS(‘bacd0 , ‘adbcd0 )
or the last letter of B is not matched:
need to compute LCS(‘bacdc0 , ‘adbc0 )
35/79
Dynamic Programming for LCS
opt[i, j], 0 ≤ i ≤ n, 0 ≤ j ≤ m: length of longest common

sub-sequence of A[1 .. i] and B[1 .. j].
if i = 0 or j = 0, then opt[i, j] = 0.
if i > 0, j > 0, then

opt[i −
 ( 1, j − 1] + 1 if A[i] = B[j]
opt[i, j] = opt[i − 1, j]
 max
 if A[i] 6= B[j]
opt[i, j − 1]
36/79
Dynamic Programming for LCS
1: for j ← 0 to m do
2: opt[0, j] ← 0
4: opt[i, 0] ← 0
6: if A[i] = B[j] then
7: opt[i, j] ← opt[i − 1, j − 1] + 1, π[i, j] ← “-”
8: else if opt[i, j − 1] ≥ opt[i − 1, j] then
9: opt[i, j] ← opt[i, j − 1], π[i, j] ←“←”
10: else
11: opt[i, j] ← opt[i − 1, j], π[i, j] ← “↑”
37/79
Example
1 2 3 4 5 6
A b a c d c a
B a d b c d a
0 1 2 3 4 5 6
0 0⊥ 0⊥ 0⊥ 0⊥ 0⊥ 0⊥ 0⊥
1 0⊥ 0← 0← 1- 1← 1← 1←
2 0⊥ 1- 1← 1← 1← 1← 2-
3 0⊥ 1↑ 1← 1← 2- 2← 2←
4 0⊥ 1↑ 2- 2← 2← 3- 3←
5 0⊥ 1↑ 2↑ 2← 3- 3← 3←
6 0⊥ 1- 2↑ 2← 3↑ 3← 4-
38/79
Example: Find Common Subsequence
1 2 3 4 5 6
A b a c d c a
B a d b c d a
0 1 2 3 4 5 6
0 0⊥ 0⊥ 0⊥ 0⊥ 0⊥ 0⊥ 0⊥
1 0⊥ 0← 0← 1- 1← 1← 1←
2 0⊥ 1- 1← 1← 1← 1← 2-
3 0⊥ 1↑ 1← 1← 2- 2← 2←
4 0⊥ 1↑ 2- 2← 2← 3- 3←
5 0⊥ 1↑ 2↑ 2← 3- 3← 3←
6 0⊥ 1- 2↑ 2← 3↑ 3← 4-
39/79
Find Common Subsequence
1: i ← n, j ← m, S ← ()
2: while i > 0 and j > 0 do
3: if π[i, j] =“-” then
4: add A[i] to beginning of S, i ← i − 1, j ← j − 1
5: else if π[i, j] =“↑” then
6: i←i−1
7: else
8: j ←j−1
9: return S
40/79
Variants of Problem
Edit Distance with Insertions and Deletions

Input: a string A
each time we can delete a letter from A or insert a letter
to A
Output: minimum number of operations (insertions or deletions) we
need to change A to B?
Example:
A = ocurrance, B = occurrence
3 operations: insert ’c’, remove ’a’ and insert ’e’
Obs. #OPs = length(A) + length(B) - 2 · length(LCS(A, B))
41/79
Variants of Problem
Edit Distance with Insertions, Deletions and Replacing

Input: a string A,
each time we can delete a letter from A, insert a letter to
A or change a letter
Output: how many operations do we need to change A to B?
Example:
A = ocurrance, B = occurrence.
2 operations: insert ’c’, change ’a’ to ’e’
Not related to LCS any more
42/79
Edit Distance (with Replacing)
opt[i, j], 0 ≤ i ≤ n, 0 ≤ j ≤ m: edit distance between A[1 .. i]

and B[1 .. j].
if i = 0 then opt[i, j] = j; if j = 0 then opt[i, j] = i.
if i > 0, j > 0, then


 opt[i − 1, j − 1] if A[i] = B[j]
 
 opt[i − 1, j] + 1
 
opt[i, j] =

 min opt[i, j − 1] + 1 if A[i] 6= B[j]
 
opt[i − 1, j − 1] + 1
 
43/79
Exercise: Longest Palindrome
Def. A palindrome is a string which reads the same backward or

forward.
example: “racecar”, “wasitacaroracatisaw”, ”putitup”
Longest Palindrome Subsequence

Input: a sequence A
Output: the longest subsequence C of A that is a palindrome.
Example:
Input: acbcedeacab
Output: acedeca
44/79
Outline

3 Knapsack Problem
8 Summary
45/79
Computing the Length of LCS
2: opt[0, j] ← 0
4: opt[i, 0] ← 0
7: opt[i, j] ← opt[i − 1, j − 1] + 1
8: else if opt[i, j − 1] ≥ opt[i − 1, j] then
9: opt[i, j] ← opt[i, j − 1]
10: else
11: opt[i, j] ← opt[i − 1, j]
Obs. The i-th row of table only depends on (i − 1)-th row.

46/79
Reducing Space to O(n + m)
Obs. The i-th row of table only depends on (i − 1)-th row.
Q: How to use this observation to reduce space?
A: We only keep two rows: the (i − 1)-th row and the i-th row.
47/79
Linear Space Algorithm to Compute Length of LCS
2: opt[0, j] ← 0
4: opt[i mod 2, 0] ← 0
7: opt[i mod 2, j] ← opt[i − 1 mod 2, j − 1] + 1
8: else if opt[i mod 2, j − 1] ≥ opt[i − 1 mod 2, j] then
9: opt[i mod 2, j] ← opt[i mod 2, j − 1]
10: else
11: opt[i mod 2, j] ← opt[i − 1 mod 2, j]
12: return opt[n mod 2, m]
48/79
How to Recover LCS Using Linear Space?
Only keep the last two rows: only know how to match A[n]
Can recover the LCS using n rounds: time = O(n2 m)
Using Divide and Conquer + Dynamic Programming:
Space: O(m + n)
Time: O(nm)
49/79
Outline

3 Knapsack Problem
8 Summary
50/79
Directed Acyclic Graphs
Def. A directed acyclic graph (DAG) is a directed graph without

(directed) cycles.
c 2 5
b d 1 3 6 8
s a
4 7
not a DAG
a DAG
Lemma A directed graph is a DAG if and only its vertices can be
topologically sorted.
51/79
Shortest Paths in DAG
Input: directed acyclic graph G = (V, E) and w : E → R.
Assume V = {1, 2, 3 · · · , n} is topologically sorted: if
(i, j) ∈ E, then i < j
Output: the shortest path from 1 to i, for every i ∈ V
2 9 5
6 3
1
8
1 2 3 5 6 5 8
9
8 2
1
4 1 7
52/79
f [i]: length of the shortest path from 1 to i

(
0 i=1
f [i] =
minj:(j,i)∈E {f (j) + w(j, i)} i = 2, 3, · · · , n
53/79
Use an adjacency list for incoming edges of each vertex i

1: f [1] ← 0 print-path(t)
2: for i ← 2 to n do 1: if t = 1 then
3: f [i] ← ∞ 2: print(1)
4: for each incoming edge (j, i) of i do 3: return
5: if f [j] + w(j, i) < f [i] then 4: print-path(π(t))
6: f [i] ← f [j] + w(j, i) 5: print(“,”, t)
7: π(i) ← j
54/79
Example
1 10
2 9 5
6 3
0 1 2 8 7 11
1 2 3 5 6 5 8
9
8 8 1 9 2
4 1 7
55/79
Variant: Heaviest Path in a Directed Acyclic Graph
Heaviest Path in a Directed Acyclic Graph

Input: directed acyclic graph G = (V, E) and w : E → R.
Assume V = {1, 2, 3 · · · , n} is topologically sorted: if
(i, j) ∈ E, then i < j
Output: the path with the largest weight (the heaviest path) from
1 to n.
f [i]: weight of the heaviest path from 1 to i

(
0 i=1
f [i] =
maxj:(j,i)∈E {f (j) + w(j, i)} i = 2, 3, · · · , n
56/79
Outline

3 Knapsack Problem
8 Summary
57/79
Matrix Chain Multiplication
Matrix Chain Multiplication

Input: n matrices A1 , A2 , · · · , An of sizes
r1 × c1 , r2 × c2 , · · · , rn × cn , such that ci = ri+1 for every
i = 1, 2, · · · , n − 1.
Output: the order of computing A1 A2 · · · An with the minimum
number of multiplications
Fact Multiplying two matrices of size r × k and k × c takes

r × k × c multiplications.
58/79
Example:
A1 : 10 × 100, A2 : 100 × 5, A3 : 5 × 50
10 × 100 100 × 5 5 × 50 10 × 100 100 × 5 5 × 50
10 · 100 · 5 100 · 5 · 50
10 × 5 = 5000 = 25000 100 × 50
10 · 5 · 50 10 · 100 · 50
10 × 50 = 2500 = 50000 10 × 50
cost = 5000 + 2500 = 7500 cost = 25000 + 50000 = 75000
(A1 A2 )A3 : 10 × 100 × 5 + 10 × 5 × 50 = 7500

A1 (A2 A3 ): 100 × 5 × 50 + 10 × 100 × 50 = 75000
59/79
Matrix Chain Multiplication: Design DP
Assume the last step is (A1 A2 · · · Ai )(Ai+1 Ai+2 · · · An )

Cost of last step: r1 × ci × cn
Optimality for sub-instances: we need to compute A1 A2 · · · Ai
and Ai+1 Ai+2 · · · An optimally
opt[i, j] : the minimum cost of computing Ai Ai+1 · · · Aj
(
0 i=j
opt[i, j] =
mink:i≤k<j (opt[i, k] + opt[k + 1, j] + ri ck cj ) i < j
60/79
Matrix Chain Multiplication: Design DP
matrix-chain-multiplication(n, r[1..n], c[1..n])

1: let opt[i, i] ← 0 for every i = 1, 2, · · · , n
2: for ` ← 2 to n do
3: for i ← 1 to n − ` + 1 do
4: j ←i+`−1
5: opt[i, j] ← ∞
6: for k ← i to j − 1 do
7: if opt[i, k] + opt[k + 1, j] + ri ck cj < opt[i, j] then
8: opt[i, j] ← opt[i, k] + opt[k + 1, j] + ri ck cj
9: π[i, j] ← k
10: return opt[1, n]
61/79
Constructing Optimal Solution
Print-Optimal-Order(i, j)
1: if i = j then
2: print(“A”i )
3: else
4: print(“(”)
5: Print-Optimal-Order(i, π[i, j])
6: Print-Optimal-Order(π[i, j] + 1, j)
7: print(“)”)
62/79
matrix A1 A2 A3 A4 A5
size 3×5 5×2 2×6 6×9 9×4
opt[1, 2] = opt[1, 1] + opt[2, 2] + 3 × 5 × 2 = 30, π[1, 2] = 1

opt[2, 3] = opt[2, 2] + opt[3, 3] + 5 × 2 × 6 = 60, π[2, 3] = 2
opt[3, 4] = opt[3, 3] + opt[4, 4] + 2 × 6 × 9 = 108, π[3, 4] = 3
opt[4, 5] = opt[4, 4] + opt[5, 5] + 6 × 9 × 4 = 216, π[4, 5] = 4
opt[1, 3] = min{opt[1, 1] + opt[2, 3] + 3 × 5 × 6,
opt[1, 2] + opt[3, 3] + 3 × 2 × 6}
= min{0 + 60 + 90, 30 + 0 + 36} = 66, π[1, 3] = 2
opt[2, 4] = min{opt[2, 2] + opt[3, 4] + 5 × 2 × 9,
opt[2, 3] + opt[4, 4] + 5 × 6 × 9}
= min{0 + 108 + 90, 60 + 0 + 270} = 198, π[2, 4] = 2,
63/79
size 3×5 5×2 2×6 6×9 9×4
opt[3, 5] = min{opt[3, 3] + opt[4, 5] + 2 × 6 × 4,

opt[3, 4] + opt[5, 5] + 2 × 9 × 4}
= min{0 + 216 + 48, 108 + 0 + 72} = 180,
π[3, 5] = 4,
opt[1, 4] = min{opt[1, 1] + opt[2, 4] + 3 × 5 × 9,
opt[1, 2] + opt[3, 4] + 3 × 2 × 9,
opt[1, 3] + opt[4, 4] + 3 × 6 × 9}
= min{0 + 198 + 135, 30 + 108 + 54, 66 + 0 + 162} = 192,
π[1, 4] = 2,
64/79
size 3×5 5×2 2×6 6×9 9×4
opt[2, 5] = min{opt[2, 2] + opt[3, 5] + 5 × 2 × 4,

opt[2, 3] + opt[4, 5] + 5 × 6 × 4,
opt[2, 4] + opt[5, 5] + 5 × 9 × 4}
= min{0 + 180 + 40, 60 + 216 + 120, 198 + 0 + 180} = 220,
opt[1, 5] = min{opt[1, 1] + opt[2, 5] + 3 × 5 × 4,
opt[1, 2] + opt[3, 5] + 3 × 2 × 4,
opt[1, 3] + opt[4, 5] + 3 × 6 × 4,
opt[1, 4] + opt[5, 5] + 3 × 9 × 4}
= min{0 + 220 + 60, 30 + 180 + 24,
66 + 216 + 72, 192 + 0 + 108}
= 234,
π[1, 5] = 2.
65/79
size 3×5 5×2 2×6 6×9 9×4
opt, π j=1 j=2 j=3 j=4 j=5

i=1 0, / 30, 1 66, 2 192, 2 234, 2
i=2 0, / 60, 2 198, 2 220, 2
i=3 0, / 108, 3 180, 4
i=4 0, / 216, 4
i=5 0, /
66/79
opt, π j=1 j=2 j=3 j=4 j=5
i=1 0, / 30, 1 66, 2 192, 2 234, 2
i=2 0, / 60, 2 198, 2 220, 2
i=3 0, / 108, 3 180, 4
i=4 0, / 216, 4
i=5 0, /
Print-Optimal-Order(1,5)
Print-Optimal-Order(1, 2)
Optimum way for multiplication: ((A1 A2 )((A3 A4 )A5 ))
67/79
Outline

3 Knapsack Problem
8 Summary
68/79
Optimum Binary Search Tree
n elements e1 < e2 < e3 < · · · < en

ei has frequency fi
goal: build a binary search tree for {e1 , e2 , · · · , en } with the
minimum accessing cost:
n
X
fi × (depth of ei in the tree)
i=1
69/79
Example: f1 = 10, f2 = 5, f3 = 3
e1 e2 e3
e2 e1 e3 e2
e3 e1
10 × 1 + 5 × 2 + 3 × 3 = 29
10 × 2 + 5 × 1 + 3 × 2 = 31
10 × 3 + 5 × 2 + 3 × 1 = 43
70/79
suppose we decided to let ek be the root
e1 , e2 , · · · , ek−1 are on left sub-tree
ek+1 , ek+2 , · · · , en are on right sub-tree
dj : depth of ej in our tree
C, CL , CR : cost of tree, left sub-tree and right sub-tree
d1 = 3, d2 = 2, d3 = 3, d4 = 4, d5 = 1,
e5
d6 = 2, d7 = 4, d8 = 3, d9 = 4,
C = 3f1 + 2f2 + 3f3 + 4f4 + f5 +
e2 e6
2f6 + 4f7 + 3f8 + 4f9
e1 e3 e8 CL = 2f1 + f2 + 2f3 + 3f4
e4 e7 e9
CR = f6 + 3f7 + 2f8 + 3f9
C = CL + CR + 9j=1 fj Z
P
71/79
C: cost of left tree
ek
CL : cost of left tree CR : cost of left tree
e1 · · · ek−1 ek+1 · · · en
n
X n
X n
X
C= f ` d` = f` (d` − 1) + f`
`=1 `=1 `=1
k−1
X n
X n
X
= f` (d` − 1) + f` (d` − 1) + f`
`=1 `=k+1 `=1
n
X
= CL + CR + f`
`=1
72/79
n
X
C = CL + CR + f`
`=1
In order to minimize C, need to minimize CL and CR respectively

opt[i, j]: the optimum cost for the instance (fi , fi+1 , · · · , fj )
n
X
opt[1, n] = min (opt[1, k − 1] + opt[k + 1, n]) + f`
k:1≤k≤n
`=1
In general, opt[i, j] =
(
0 if i = j + 1
mink:i≤k≤j opt[i, k − 1] + opt[k + 1, j] + j`=i f`
P
if i ≤ j
73/79
1: f sum[0] ← 0
2: for i ← 1 to n do f sum[i] ← f sum[i − 1] + fi
. f sum[i] = ij=1 fj
P
3: for i ← 0 to n do opt[i + 1, i] ← 0
4: for ` ← 1 to n do
5: for i ← 1 to n − ` + 1 do
6: j ← i + ` − 1, opt[i, j] ← ∞
7: for k ← i to j do
8: if opt[i, k − 1] + opt[k + 1, j] < opt[i, j] then
9: opt[i, j] ← opt[i, k − 1] + opt[k + 1, j]
10: π[i, j] ← k
11: opt[i, j] ← opt[i, j] + f sum[j] − f sum[i − 1]
74/79
Printing the Tree
Print-Tree(i, j)
1: if i > j then
2: return
3: else
4: print(’(’)
5: Print-Tree(i, π[i, j] − 1)
6: print(π[i, j])
7: Print-Tree(π[i, j] + 1, j)
8: print(’)’)
75/79
Outline

3 Knapsack Problem
8 Summary
76/79
Dynamic Programming
77/79
Comparison with greedy algorithms
Greedy algorithm: each step is making a small progress towards
constructing the solution
Dynamic programming: the whole solution is constructed in the
last step
Comparison with divide and conquer

Divide and conquer: an instance is broken into many independent
sub-instances, which are solved separately.
Dynamic programming: the sub-instances we constructed are
overlapping.
78/79
Definition of Cells for Problems We Learnt
Weighted interval scheduling: opt[i] = value of instance defined

by jobs {1, 2, · · · , i}
Subset sum, knapsack: opt[i, W 0 ] = value of instance with items
{1, 2, · · · , i} and budget W 0
Longest common subsequence: opt[i, j] = value of instance
defined by A[1..i] and B[1..j]
Shortest paths in DAG: f [v] = length of shortest path from s to v
Matrix chain multiplication, optimum binary search tree:
opt[i, j] = value of instances defined by matrices i to j
79/79

Dynamic Programming: CSE 431/531: Algorithm Analysis and Design (Spring 2022)

Uploaded by

Copyright:

Available Formats

Dynamic Programming: CSE 431/531: Algorithm Analysis and Design (Spring 2022)

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Dynamic Programming: CSE 431/531: Algorithm Analysis and Design (Spring 2022)

Uploaded by

Copyright:

Available Formats

CSE 431/531: Algorithm Analysis and Design (Spring 2022)

Store each F [i] for future use.

1 Weighted Interval Scheduling

Optimum value = 220

Q: Which job is safe to schedule?

Q: The value of optimal solution that does not contain i?

Q: The value of optimal solution that contains job i?

A: vi + opt[pi ], pi = the largest j such that fj ≤ si

Q: The value of optimal solution that does not contain i?

Q: The value of optimal solution that contains job i?

A: vi + opt[pi ], pi = the largest j such that fj ≤ si

Recursion for opt[i]:

Recursion for opt[i]:

Recursion for opt[i]:

opt[0] = 0, opt[1] = 80, opt[2] = 100

1: sort jobs by non-decreasing order of finishing times

Running time sorting: O(n lg n)

1: sort jobs by non-decreasing order of

Break up a problem into many overlapping sub-problems

1 Weighted Interval Scheduling

Motivation: you have budget W , and want to buy a subset of

Q: Does candidate algorithm always produce optimal solutions?

A: No. W = 100, n = 3, w = (51, 50, 50).

Q: What if we change “non-increasing” to “non-decreasing”?

A: No. W = 100, n = 3, w = (1, 50, 50)

Consider the instance: i, W 0 , (w1 , w2 , · · · , wi );

Q: The value of the optimum solution that does not contain i?

Q: The value of the optimum solution that contains i?

Consider the instance: i, W 0 , (w1 , w2 , · · · , wi );

Running time is O(nW )

Use hash map for opt

1 Weighted Interval Scheduling

Motivation: you have budget W , and want to buy a subset of

opt[i, W 0 ]: the optimum value when budget is W 0 and items are

Input: integer bounds W > 0, Z > 0,

1 Weighted Interval Scheduling

Def. Given two sequences A[1 .. n] and C[1 .. t] of letters, C is

Exercise: how to check if sequence C is a subsequence of A?

Applications: edit distance (diff), similarity of DNAs

Goal of LCS: find a maximum-size non-crossing matching between

opt[i, j], 0 ≤ i ≤ n, 0 ≤ j ≤ m: length of longest common

Edit Distance with Insertions and Deletions

Obs. #OPs = length(A) + length(B) - 2 · length(LCS(A, B))

Edit Distance with Insertions, Deletions and Replacing

Not related to LCS any more

opt[i, j], 0 ≤ i ≤ n, 0 ≤ j ≤ m: edit distance between A[1 .. i]

Def. A palindrome is a string which reads the same backward or

example: “racecar”, “wasitacaroracatisaw”, ”putitup”

Longest Palindrome Subsequence

1 Weighted Interval Scheduling

Obs. The i-th row of table only depends on (i − 1)-th row.

Obs. The i-th row of table only depends on (i − 1)-th row.

Q: How to use this observation to reduce space?

1 Weighted Interval Scheduling

Def. A directed acyclic graph (DAG) is a directed graph without

f [i]: length of the shortest path from 1 to i

Use an adjacency list for incoming edges of each vertex i

Shortest Paths in DAG