the greedy approachgsyoung/cs5300/cs530 hybrid notes/cs530 … · greedy 3 young cs 530 adv. algo....
TRANSCRIPT
Greedy
1
YoungCS 530 Adv. Algo.
Greedy 1
The Greedy Approach
YoungCS 530 Adv. Algo.
Greedy 2
General idea:
Given a problem with n inputs, we are required to obtain a subset that maximizes or minimizes a given objective function subject to some constraints.
Feasible solution — any subset that satisfies some constraints
Optimal solution — a feasible solution that maximizes or minimizes the objective function
Greedy
2
YoungCS 530 Adv. Algo.
Greedy 3
procedure Greedy (A, n)begin
solution ← Ø;for i ← 1 to n do
x ← Select (A); // based on the objective // function
if Feasible (solution, x), then solution ← Union (solution, x);
end;Select: A greedy procedure, based on a given objective function, which selects input from A, removes it and assigns its value to x.
Feasible: A boolean function to decide if x can be included into solution vector (without violating any given constraint).
YoungCS 530 Adv. Algo.
Greedy 4
About Greedy method
The n inputs are ordered by some selection procedure which is based on some optimization measures.
It works in stages, considering one input at a time. At each stage, a decision is made regarding whether or not a particular input is in an optimal solution.
Greedy
3
YoungCS 530 Adv. Algo.
Greedy 5
1. Minimum Spanning Tree ( For Undirected Graph)
The problem:1) Tree
A Tree is connected graph with no cycles.
2) Spanning Tree
A Spanning Tree of G is a tree which contains all verticesin G.Example:
G:
YoungCS 530 Adv. Algo.
Greedy 6
b) Is G a Spanning Tree?
Key: Yes Key: No
Note: Connected graph with n vertices and exactly n – 1 edges is Spanning Tree.
3) Minimum Spanning Tree
Assign weight to each edge of G, then Minimum Spanning Tree is the Spanning Tree with minimum total weight.
Greedy
4
YoungCS 530 Adv. Algo.
Greedy 7
Example:a) Edges have the same weight
G: 1
2 3
4 5 6 7
8
YoungCS 530 Adv. Algo.
Greedy 8
DFS (Depth First Search)
1
2 3
4
5
6
7
8
Greedy
5
YoungCS 530 Adv. Algo.
Greedy 9
BFS (Breadth First Search)
1
2 3
4 5 6 7
8
YoungCS 530 Adv. Algo.
Greedy 10
b) Edges have different weights
G:
DFS
1 2
3
45
6
16
19
21
33
11
14
18
10
5
6
1
5
2
3
4
6
16
5
10
143378
331410516=
++++=Cost
Greedy
6
YoungCS 530 Adv. Algo.
Greedy 11
BFS
Minimum Spanning Tree (with the least total weight)
1 2 3
45
6
16
1921
5
6
6765211916
=++++=Cost
1 2
3
4 56
165
611
18 5618116516
=++++=Cost
YoungCS 530 Adv. Algo.
Greedy 12
Algorithms:1) Prim’s Algorithm (Minimum Spanning Tree)
Basic idea:Start from vertex 1 and let T ← Ø (T will contain all edges in the S.T.); the next edge to be included in T is the minimum cost edge(u, v), s.t. u is in the tree and v is not.
Example: G1 2
3
45
6
16
19
21
33
11
14
18
10
5
6
Greedy
7
YoungCS 530 Adv. Algo.
Greedy 13
1
2 5 6
1619 21
1 2
5 6 3 64
19 21 5 6 11
1 1 2
1 2 3
5 6 4 6 4
1921 6 11 10
1 2 3 4
5 6 6 5 6
19 21 11 18 14
1 2 3 1 2 3
4
S.T. S.T.
S.T.S.T.
(Spanning Tree)
YoungCS 530 Adv. Algo.
Greedy 14
1 2 3 4 6
5 5 5
19 18 33
1 2 3
4 6
S.T.
(n – # of vertices, e – # of edges)It takes O(n) steps. Each step takes O(e) and e ≤ n(n-1)/2Therefore, it takes O(n3) time.
With clever data structure, it can be implemented in O(n2).
).( 2nO⇒
1 2 3
4 6
5
S.T.
Cost = 16+5+6+11+18=56Minimum Spanning Tree
Greedy
8
YoungCS 530 Adv. Algo.
Greedy 15
Example:
Step 1: Sort all of edges(1,2) 10 √(3,6) 15 √(4,6) 20 √(2,6) 25 √(1,4) 30 × reject Qcreate cycle(3,5) 35 √
1 2
3
5
4
6
1050
35
40
55
15
4525
30
20
2) Kruskal’s AlgorithmBasic idea:
Don’t care if T is a tree or not in the intermediate stage, as long as the including of a new edge will not create a cycle, we include the minimum cost edge
YoungCS 530 Adv. Algo.
Greedy 16
Step 2: T
1 2
1 2 3 6
1 2 3 6 4
1 2 3 6 4
1 2 3 6 4
5
Greedy
9
YoungCS 530 Adv. Algo.
Greedy 17
How to check:
adding an edge will create a cycle or not?
If Maintain a set for each group(initially each node represents a set)Ex: set1 set2 set3
∴ new edge
from different groups ⇒ no cycle created
Data structure to store sets so that:i. The group number can be easily found, and
ii. Two sets can be easily merged
1 2 3 6 4 5
2 6
YoungCS 530 Adv. Algo.
Greedy 18
While (T contains fewer than n-1 edges) and (E ≠ ∅ ) doBegin
Choose an edge (v,w) from E of lowest cost;Delete (v,w) from E;If (v,w) does not create a cycle in Tthen add (v,w) to T else discard (v,w);
End;
With clever data structure, it can be implemented in O(e log e).
Kruskal’s algorithm
Greedy
10
YoungCS 530 Adv. Algo.
Greedy 19
So, complexity of Kruskal is )(eLogeO
LognLognLogenne 22
)1( 2 =≤⇒−
≤Q
)()( eLognOeLogeO =⇒
3) Comparing Prim’s Algorithm with Kruskal’s Algorithm
i. Prim’s complexity is
ii. Kruskal’s complexity is
if G is a complete (dense) graph, Kruskal’s complexity is .
if G is a sparse graph, Kruskal’s complexity is
)( 2nO
)(eLognO
)( 2LognnO
).(nLognO
YoungCS 530 Adv. Algo.
Greedy 20
2. Dijkstra’s Algorithm for Single-Source Shortest PathsThe problem: Given directed graph G = (V, E),
a weight for each edge in G,a source node v0,
Goal: determine the (length of) shortest paths from v0 to all the remaining vertices in G
Def: Length of the path: Sum of the weight of the edgesObservation:
May have more than 1 paths between w and x (y and z)But each individual path must be minimal length
(in order to form an overall shortest path form v0 to vi )
V0 Vi
w x y z
shortest paths from v0 to vi
Greedy
11
YoungCS 530 Adv. Algo.
Greedy 21
1 if shortest path (v, w) is defined0 otherwises(w) =
jjDist ∀)( in the vertex set V= the length of the shortest path from v to j
ijFrom =)( if i is the predecessor of jalong the shortest path from v to j
Notation
cost adjacency matrix Cost, ∀1 ≤ a,b ≤ V
Cost (a, b) = cost from vertex i to vertex j if there is a edge0 if a = b∞ otherwise
YoungCS 530 Adv. Algo.
Greedy 22
Example:V0 V1
V2 V3
V4
V5
50 10
2010
15
15
20
3
35
30
45
a) Cost adjacent matrix
∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞
03030
3502015020
101504510500
5
4
3
2
1
0
543210
vvvvvv
vvvvvv
Greedy
12
YoungCS 530 Adv. Algo.
Greedy 23
b) Steps in Dijkstra’s Algorithm
1. Dist (v0) = 0, From (v0) = v0 2. Dist (v2) = 10, From (v2) = v0
V0 V1
V2 V3
V4
V5
50 10
315
2035
30
1520 10
45
V0
V2
V1
V3
V4
V5
50 10
2010
15
15
20 35
30
3
45
YoungCS 530 Adv. Algo.
Greedy 24
3. Dist (v3) = 25, From (v3) = v2 4. Dist (v1) = 45, From (v1) = v3
V0
V2 V3
V1 V4
V5
50 10
45
20 10
15
15
2035
3
V0
V2
V1
V3
V4
V5
50
45
10
20 1015
15
20
35
30
3
30
Greedy
13
YoungCS 530 Adv. Algo.
Greedy 25
5. Dist (v4) = 45, From (v4) = v0 6. Dist (5) = ∞
V0 V1
V2 V3
V4
V5
45
50 10
20 10
15
20
15
35
30
3
V0 V1
V2 V3
V4
V5
45
50 10
20 10
15
20
15
35
30
3
YoungCS 530 Adv. Algo.
Greedy 26
c) Shortest paths from source v0
v0 → v2 → v3 → v1 45
v0 → v2 10
v0 → v2 → v3 25
v0 → v4 45
v0 → v5 ∞
Greedy
14
YoungCS 530 Adv. Algo.
Greedy 27
Dijkstra’s algorithm:
procedure Dijkstra (Cost, n, v, Dist, From)// Cost, n, v are input, Dist, From are output
beginfor i ← 1 to n do
for num ← 1 to (n – 1) do choose u s.t. s(u) = 0 and Dist(u) is minimum;
for all w with s(w) = 0 doif
end;
(cont. next page)
;)();,()(
;0)(
viFromivCostiDist
is
←←
←
;1)( ←vs
;1)( ←us
;)();,()()(
uwFromwuCostuDistwDist
←+←
))(),()(( wDistwuCostuDist <+
YoungCS 530 Adv. Algo.
Greedy 28
Ex: 1
25
34
10 50
10
100 30
20
50
5
a) Cost adjacent matrix
∞∞∞∞∞∞∞∞∞∞∞∞
010020
50050
1010030500
54321
0504030201
Greedy
15
YoungCS 530 Adv. Algo.
Greedy 29
b) Steps in Dijkstra’s algorithm
1. Dist (1) = 0, From (1) = 1 2. Dist (5) = 10, From (5) = 1
1
25
34
10 50
10
100 30
20
50
5
50
1
25
34
10 50
10
100 30
205
YoungCS 530 Adv. Algo.
Greedy 30
3. Dist (4) = 20, From (4) = 5 4. Dist (3) = 30, From (3) = 11
25
34
10 50
10
100 30
20
50
5
1
25
34
10 50
10
100 30
20
50
5
5. Dist (2) = 35, From (2) = 31
25
34
10 50
10
100 30
20
50
5
Shortest paths from source 1
1 → 3 → 2 35
1 → 3 30
1 → 5→ 4 20
1 → 5 10
Greedy
16
YoungCS 530 Adv. Algo.
Greedy 31
3. Optimal Storage on Tapes
The problem:
Given n programs to be stored on tape, the lengths of these n programs are l1, l2 , . . . , ln respectively.Suppose the programs are stored in the order of i1, i2 , . . . , in
Let tj be the time to retrieve program ij.
Assume that the tape is initially positioned at the beginning.
tj is proportional to the sum of all lengths of programs stored infront of the program ij.
YoungCS 530 Adv. Algo.
Greedy 32
The goal is to minimize MRT (Mean Retrieval Time),
i.e. want to minimize
∑=
n
jjt
n 1
1
∑∑= =
n
j
j
kkil
1 1
Ex:There are n! = 6 possible orderings for storing them.
)3,10,5(),,(,3 321 == llln
order total retrieval time MRT
123456
1 2 31 3 22 1 32 3 13 1 23 2 1
5+(5+10)+(5+10+3)=385+(5+3)+(5+3+10)=31
10+(10+5)+(10+5+3)=4310+(10+3)+(10+3+5)=413+(3+5)+(3+5+10)=293+(3+10)+(3+10+5)=34
38/331/343/341/329/334/3
Note: The problem can be solved using greedy strategy,just always let the shortest program goes first.
( Can simply get the right order by using any sorting algorithm)
Smallest
Greedy
17
YoungCS 530 Adv. Algo.
Greedy 33
Analysis:
Try all combination: O( n! )
Shortest-length-First Greedy method: O (nlogn)
Shortest-length-First Greedy method: Sort the programs s.t. and call this ordering L.
Next is to show that the ordering L is the best
Proof by contradiction:Suppose Greedy ordering L is not optimal, then there exists some other permutation I that is optimal.I = (i1, i2, … in) ∃ a < b, s.t. (otherwise I = L)
ba ii ll >
nlll ≤≤≤ . . .21
YoungCS 530 Adv. Algo.
Greedy 34
Interchange ia and ib in and call the new list I′ :I
I′
… ia ia+1 ia+2 … ib …
x
… ib ia+1 ia+2 … ia …
xIn I′, Program ia+1 will take less (lia
- lib) time than in I to be retrieved
In fact, each program ia+1 , …, ib-1 will take less (lia- lib
) time For ib, the retrieval time decreases x + liaFor ia, the retrieval time increases x + lib
)(0))((
)()())(1()()(
→←>−−=
+−++−−−=′−
ba
baba
ii
iiii
llab
lxlxllabItotalRTItotalRT
Therefore, greedy ordering L is optimal Contradiction!!
SWAP
Greedy
18
YoungCS 530 Adv. Algo.
Greedy 35
4. Knapsack Problem
The problem:Given a knapsack with a certain capacity M,n objects, are to be put into the knapsack, each has a weight and a profit if put in the knapsack .
The goal is find where
s.t. is maximized and
Note: All objects can break into small piecesor xi can be any fraction between 0 and 1.
nwww ,,, 21 L
nppp ,,, 21 L
),,,( 21 nxxx L 10 ≤≤ ix
∑=
n
iii xp
1∑=
≤n
iii Mxw
1
YoungCS 530 Adv. Algo.
Greedy 36
Example:
)15,24,25(),,()10,15,18(),,(
203
321
321
==
==
pppwww
Mn
Greedy Strategy#1: Profits are ordered in nonincreasing order (1,2,3)
2.2801515224125
)0,152,1(),,(
3
1
321
=×+×+×=
=
∑=i
ii xp
xxx
Greedy
19
YoungCS 530 Adv. Algo.
Greedy 37
Greedy Strategy#2: Weights are ordered in nondecreasing order (3,2,1)
311153224025
)1,32,0(),,(
3
1
321
=×+×+×=
=
∑=i
ii xp
xxx
Greedy Strategy#3: p/w are ordered in nonincreasing order (2,3,1)
5.312115124025
)21,1,0(),,(
)1,3,2(
5.11015
6.11524
4.11825
3
1
321
3
3
2
2
1
1
=×+×+×=
=
⇒
==
==
==
∑=i
ii xp
xxx
wpwpwp
Optimal solution
YoungCS 530 Adv. Algo.
Greedy 38
Analysis:Sort the p/w, such that Show that the ordering is the best.
Proof by contradiction:Given some knapsack instanceSuppose the objects are ordered s.t.
let the greedy solution be
Show that this ordering is optimalCase1: it’s optimalCase2: s.t.
where
n
n
wp
wp
wp
≥≥≥ L2
2
1
1
n
nwp
wp
wp
≥≥≥ L2
2
1
1
),,( 21 nxxxX L=
)1,,1,1( L=X)0,,0,,,1,1( LL jxX = ∑
=
=n
iii Mxw
1
10 ≤≤ jx
Greedy
20
YoungCS 530 Adv. Algo.
Greedy 39
Assume X is not optimal, and then there exists
s.t. and Y is optimal
examine X and Y, let yk be the 1st one in Y that yk ≠ xk.
),,,( 21 nyyyY L=
∑∑==
>n
iii
n
iii xpyp
11
),,,,,,( 121 nkk xxxxxX LLLL −=
),,,,,,( 121 nkk yyyyyY LLLL −=
Now we increase yk to xk and decrease as many of as necessary, so that the capacity is still M.
),,( 1 nk yy L+
same yk ≠ xk⇒ yk < xk
YoungCS 530 Adv. Algo.
Greedy 40
Let this new solution be
where
),,,( 21 nzzzZ L=kixz ii ≤≤∀= 1
∑+=
⋅−=⋅−n
kiiiikkk wzywyzand
1
)()(
∑∑∑∑
∑∑∑
=+===
+===
=⋅
⋅−−⋅−+≥
⋅≤∴
+≥∀≥
⋅−−⋅−+=
n
iii
k
kn
kiiiikkk
n
iii
n
iii
ik
ki
i
i
k
k
n
kiiiikkk
n
iii
n
iii
ypwpwzywyzypzp
wwpp
kiwp
wp
pzypyzypzp
1111
111
)()(
)(
1
)()(
Q
0
Greedy
21
YoungCS 530 Adv. Algo.
Greedy 41
So,if
else (Repeat the same process. At the end, Y can be transformed into X.⇒ X is also optimal.Contradiction! )
)()()( →←> yprofitzprofit
)()( yprofitzprofit =
→←