1. Longest Common Subsequence
Dynamic Programming
Longest Common Subsequence

2. Dynamic Programming(DP)
A powerful design technique for optimization problems
Related to divide and conquer
However, due to the nature of DP problems, standard divide-and-conquer solution are not efficient

3. Dynamic Programming(DP)
Main question: How to set up the subproblem structure?
For DP to be applicable to an optimization problem
Optimal substructure: for the global problem to be solved optimally, each subproblem should be solved optimally
Polynomially many subproblems
Overlapping subproblems

4. Longest Common Subsequence (LCS)
Application: searching for a substring or pattern in a large piece of text
Not necessarily exact text but something similar
Method for measuring degree of similarity: LCS

5. Longest Common Subsequence (LCS)
Given two sequences x[1 . . m] and y[1 . . n], find a longest subsequence common to them both.
x: A B C B D A B
y: B D C A B A

6. Longest Common Subsequence (LCS)
Given two sequences x[1 . . m] and y[1 . . n], find a longest subsequence common to them both.
x: A B C B D A B
y: B D C A B A
LCS(x,y) = BCBA

7. LCS
Not always unique
X : A B C
Y: B A C

8. LCS
Not always unique
X : A B C
Y: B A C

9. Brute Force Solution?
Check every subsequence of x[1 . . m] to see if it is also a subsequence of y[1 . .n]

10. Brute Force Solution?
Check every subsequence of x[1 . . m] to see if it is also a subsequence of y[1 . .n]
Analysis
• Checking = O(n) time per subsequence.
• subsequences of x(each bit-vector of length m determines a distinct subsequence of x).
Thus, Exponential time.

11. DP Formulation for LCS Subproblems? Consider all pairs of prefixes
A prefix of a sequence is just an initial string of characters
Let
denote the prefix of X with i characters
Let denote the empty sequence
x: A B C B D A B

12. DP Formulation Compute the LCS for every possible pair of prefixes
C[i,j]: length of the LCS of and
Then, C[m,n]: length of LCS of X and Y
x: A B C B D A B
y: B D C A B A

13. Recursive formulation
C[i,0] = C[0,j] = 0 // base case
How can I compute C[i,j] using solutions to subproblems?

14. Recursive formulation
Two cases
Last characters match: if X[i] = Y[j]
LCS must end with this same character
x: A B C B D A B
y: B D C A B A i X A j Y A

15. Recursive formulation
Two cases
Last characters match: if X[i] = Y[j]
LCS must end with this same character
C[i,j] = C[i-1, j-1] + 1
LCS ( ) = LCS( ) + X[i] i X A j Y A

16. Recursive formulation
(2) Last characters DO NOT match:
LCS( ) //skip x[i] LCS( ) //skip Y[j]
C[i,j] = max (C[i, j-1], C[i-1,j]) B A i j X Y B A i j X Y B A i j X Y

17. A recursive algorithm LCS(x, y, i, j) { if x[i] = y[ j ]
then c[i, j ] ←LCS(x, y, i–1, j–1) + 1
else c[i, j ] ←max{LCS(x, y, i–1, j), LCS(x, y, i, j–1) }
Worst-case:x[i] ≠y[ j ], in which case the algorithm evaluates two subproblems, each with only one parameter decremented.

18. Recursion tree Height = O(m+n) Running time = potentially exponential!
3,4
2,4
3,3
1,4
2,3
3,2
Height = O(m+n)
Running time = potentially exponential!

19. BUT, we keep recomputing the same subproblems
3,4
2,4
3,3
1,4
2,3
2,3
3,2

20. Overlapping subproblems
A recursive solution contains a “small” number of distinct subproblems repeated many times
The number of distinct LCS subproblems for two strings of lengths m and n is only mn

21. Store the solutions of subproblems
After computing a solution to a subproblem, store it in a table.
C[0..m,0..n] that stores lengths of LCS
Keep also a helper array B[0..m,0..n] to store some pointers to extract the LCS later

22. O(mn) LCS(x[1..m],y[1..n]) { int C[0..m,0..n]; int B[0..m,0..n]
for i=0 to m
C[i,0] = 0; B[i,0] = UP
for j=0 to n
C[0,j] = 0; B[0,j] = LEFT
for i=1 to m
for j=1 to n
if x[i] == y[j]
C[i,j]=c[i-1, j-1] +1; B[i,j] = DIAG;
else if (C[i-1,j] >= C[i, j-1])
C[i,j]=C[i-1, j]; B[i,j] = UP;
else C[i,j]=C[i, j-1]; B[i,j] = LEFT;
return C[m,n]; }
O(mn)

23. Compute tables bottom-upY Y B A C D B D C B B A X X C D B
Length Table: C
Tables C and B

24. Compute tables bottom-upY Y B A C D B D C B B A X X C D B
Length Table: C
Tables C and B
Start from B[m,n], follow pointers
Extract entries with DIAG ( )
LCS for above example: BCB

25. ExtractLCS(B, X, i, j) { //initially called with (B, X, m, n)
if i==0 OR j== 0 return;
if B[i,j] == DIAG
ExtractLCS(B, X, i-1, j-1);
print X[i]
else if B[i,j] == UP
extractLCS(B, X, i-1, j);
else // LEFT
extractLCS(B, X, i, j-1) }

26. Example of LCF